Chapter 4E: ARC as Constitutional Remedy

April 21, 2026

Chapter 4E: ARC as Constitutional Remedy

Analyzes the At-Risk Coefficient (ARC), developed by Kevin Hopper (2025), as a measurement framework for cumulative student risk grounded in TEC §29.081 statutory criteria. Tests whether the frozen statutory compensatory weight of 0.20 (unchanged since 1984) meets the constitutional standard of an 'efficient system.' Presents three recalibrated regression models (ARC-S Dropout R²=0.45, ARC-S Absenteeism R²=0.31, ARC-P Absenteeism R²=0.44), computes empirically-derived funding weights, and quantifies a statewide funding gap of $2.28 billion. Examines vertical equity through concentration analysis (Spearman ρ=-0.24, Kakwani=0.09) and tests whether TIA incentive funding follows student need. Calls for the 90th Legislature to order an updated Accountable Costs study using the ARC framework.

The Texas compensatory education weight has been 0.20 since House Bill 72 (68th Legislature, 1984). Section 16.151 of HB 72 mandated that the State Board of Education conduct a cost study to determine whether this weight was adequate. The resulting Accountable Costs study (Texas Education Agency & Accountable Costs Advisory Committee, 1986), overseen by Deputy Commissioner Lynn M. Moak, found the 0.20 weight roughly appropriate for the at-risk population as then defined. The committee also recommended increasing the bilingual education weight from 0.10 to 0.26 based on empirical cost analysis. The Legislature never enacted that recommendation. The committee recommended that the state assume responsibility for equalized construction funding. The Legislature created IFA and EDA instead of implementing equalized state funding.

The at-risk definition in 1984 encompassed a narrower population than today's 15 criteria under TEC §29.081(d). Homelessness, foster care, chronic absenteeism, and pregnancy/parenting were not factors in 1984. The student demographic profile has shifted substantially in four decades. No updated cost study has been conducted since 1986, and the 0.20 weight remains frozen in TEC §48.104.

The At-Risk Coefficient (ARC) was developed by Kevin Hopper (2025) as a campus-level measure of cumulative student risk, allowing a comparison across campuses in order to guide resource allocation decisions locally. Initial factor weights were drawn from a meta-analysis of school absenteeism and dropout risk factors (Gubbels, van der Put, & Assink, 2019). ARC weights each risk factor through a regression with the outcome of interest, unlike the simple at-risk percentage, which counts any student meeting at least one of the 15 statutory criteria without distinguishing between a campus where 60% of students face one risk factor and a campus where 60% face four compounding factors. Eight risk factors map to 12 of the 15 TEC §29.081 criteria. Five criteria are excluded because TEA does not report campus-level aggregate counts in its standard reporting (parole/probation, prior dropout, residential placement, incarceration, and dropout recovery enrollment).

To test whether the meta-analysis weights hold for Texas, Ordinary Least Squares (OLS) regression models were calculated using 2023-24 campus-level PEIMS data. The elementary model (N = 4,594, R² = 0.40) identifies economic disadvantage (β* = 0.40) and academic underperformance (β* = 0.36) as dominant predictors of chronic absenteeism. The middle school model (N = 1,664, R² = 0.46) finds academic underperformance (β* = 0.34) and discipline/DAEP placement (β* = 0.33) as co-dominant predictors. At middle school, discipline reaches β* = 0.33. The high school model (N = 1,565, R² = 0.32) finds academic underperformance at β* = 0.45, to be the largest single predictor of chronic absenteeism at any school level. A dropout model was also tested for high school (N = 2,087, R² = 0.45). It produced wrong-sign coefficients. Economic disadvantage came back at β* = -0.034, not significant, predicting lower dropout after controlling for other factors. The problem is the data, not the model. 780 of 2,087 campuses (37.4%) report exactly 0.0% dropout because FERPA suppresses any count below 10, flooding the distribution with artificial zeros.

An unmasked validation using Austin ISD Student Information System data (108 campuses, 2024-25, 0% FERPA masking) tested the three-bucket model. Elementary R² rose from 0.40 to 0.73. Middle school R² rose from 0.46 to 0.74. High school R² rose from 0.32 to 0.89 using a reduced 3-predictor model, because the full model had too few high schools (N = 12) to support all predictors. FERPA masking accounts for the gap between statewide and AISD model performance. A statewide ARC computed from unmasked PEIMS data would approach the AISD performance levels.

The regression analysis reveals relative variation in need across districts. A campus with high risk factor concentrations generates a larger predicted outcome value, indicating greater compensatory need at the campus. The funding weight for each district is calculated as the ratio of its predicted compensatory need to its basic allotment enrollment. The statewide distribution of these empirical weights reveals the highest-need districts require empirical weights 2 to 3 times the frozen statutory value. Standardized beta coefficients (standardized betas) allow comparison of relative predictor importance across factors measured on different scales. R² measures how much the whole model explains (40% to 46% at the statewide level, 73% to 89% with unmasked data in Austin ISD). On a standardized scale, economic disadvantage (β* = 0.40 at elementary) and academic underperformance (β* = 0.36) contribute nearly identical predictive power. Both drive chronic absenteeism at roughly the same magnitude. The compensatory education allotment ignores this. It pays 0.20 per economically disadvantaged student whether that student attends a campus where poverty is the only concentrated risk factor or a campus where poverty, discipline, homelessness, and ELL status compound simultaneously.

Three equity indicators test the relationship between ARC scores and per-pupil funding across 1,201 districts. Spearman's rho ranks districts by need, then by funding, and checks whether the rankings align. The ρ = -0.24 (p < 0.001), meaning higher-need districts receive less, indicates the rankings do not align. The Concentration Index asks where funding pools along the need spectrum. At 0.27, funding pools among lower-need districts. The Kakwani Index subtracts the Gini coefficient of need from the Concentration Index. At 0.09, the funding distribution is more unequal than the need distribution itself. All three point the same direction. A Spearman of ρ = -0.24 (p < 0.001, n = 1,201 districts) shows higher-risk districts receive less per-pupil funding. A Concentration Index of 0.27 indicates funding concentrates among lower-need districts. A Kakwani Index of 0.09 reveals the funding distribution is more unequal than the need distribution (Texas Education Agency, 2024). The statewide funding gap between the frozen 0.20 weight and empirically derived weights totals $2.28 billion. The highest-ARC districts receive no additional recognition for compounding risk burdens. The TIA analysis from Section 4B shows that incentive funding compounds the problem.

SQ4 asked whether the ARC model provides a constitutionally adequate alternative to the frozen 0.20 weight. The three-bucket regression produces empirical weights responsive to campus-level variation in cumulative risk. The 1986 Accountable Costs committee established the precedent that empirical evidence should drive compensatory weights. The data presented here provides the measurement framework the next cost study requires. Determining the actual dollar cost of mitigating each risk factor requires a legislative cost study, the approach the 1986 committee recommended, and the requirement that HB 3 eventually removed from statute. The ARC model identifies where the need is greatest and quantifies how much it varies across the district, the campus, or even the student level. Current statute underfunds the regression-derived allotment by $2.28 billion statewide.