Background: The overall goal of the No Child Left Behind Act (NCLB) of 2001 is to close, by the end of the 2013-2014 academic year, "the achievement gap between high- and low- performing students, especially the achievement gap between minority and non-minority students and, between disadvantaged children and their more advantaged peers" (NCLB, 2001, Sec. 1001[3]). Under the federal NCLB mandates, adequate yearly progress (AYP) targets must be set for the entire period from 2002 to 2014 in order to ensure that all students and all schools eventually meet the content and performance standards adopted in their respective states. It was within this context that the Texas Education Agency (TEA) launched its Texas Assessment of Knowledge and Skills (TAKS) in spring 2003 to improve its accountability system. The accountability provisions in NCLB clearly refer to two demographic variables underlying the current inequity in public education: economic disadvantage and race/ethnicity. It is obvious that the essence of accountability, according to the NCLB, is accountability for subgroups, particularly subgroups that have historically been disadvantaged by their low income and minority statuses. It is therefore important to investigate the extent to which student performance on the 2002-2003 TAKS was determined by economic disadvantage and minority status, so that the Beaumont ISD Superintendent of Schools, School Board members, and the cabinet may have a clear baseline picture by which it can judge how well Beaumont Independent School District schools and students will be leveling the playing field from 2002 up to 2014 to ensure educational equity. Purpose: The objective of the present study was to examine the impact of three demographic variables: poverty, ethnicity, and gender on the risk of a student failing to meet the TAKS reading proficiency standards in 2003. Research Design: Purposeful with four grade levels (3rd, 5th, 8th, and 10th) and three research questions. Study Sample and Setting: Students were drawn from all of the 29 elementary and secondary schools Beaumont ISD. There are 24 (16 elementary, two high, and six middle schools) school-wide Title 1 campuses in BISD. The total sample consisted of 75 teachers (11 male and 64 females) with average cumulative length of service as 12.89 years (minimum was two and maximum was 37 years); and 35%, 15%, and 50% of these teachers were African, Hispanic, and Caucasian Americans respectively. There were a total of 6,112 students in this study: 1,648 third graders, 1,560 fifth graders, 1,502 eight graders, and 1,402 tenth graders. Intervention and Control/Comparison Condition: None. Data Collection and Analysis: Data for this present investigation were collected from the district's database and state's achieves at the campus level namely the Texas state Academic Excellence Indicator System (AEIS) database of the State of TEA for the 2001-2003 school years. Within the TEA database are information about individual students and teachers and campuses. The dependent variable in this study is the binary variable of pass/fail (pass = 1, fail = 0). The event of failure (0) is modeled in logistic regression. Findings: The three-predictor model can correctly classify 65.0%, 64.8%, 64.5%, and 64.8% of the students into the "pass" or "fail" group at grades 3, 5, 8 and 10 respectively. That is, without any consideration to academic capability, roughly 65% of the students' TAKS reading results could be correctly placed. Conclusions: As expected, girls have a significantly lower failure rate than boys in reading across the grade levels, with statistically significant odds ratios of 0.73, 0.61, 0.54 and 0.49 for grades 3, 5, 8 and 10 respectively. The present study is limited by the absence of many other demographic variables that might conceivably have contributed to the failure rates on the 2002-2003 TAKS reading tests. It also faced the methodological challenge of how to include numerous smaller subgroups into the analyses. The predicted probabilities of failure used in classifying the students into the predicted pass and fail groups may be optimistically biased because the predicted results and the actual results are from the same data. Validations using 2004-2005 and 2005-2006 TAKS data are under consideration. (Contains 4 tables.)