Background: Nationally, about 15% of students -- about 7.3 million -- receive special education services. Each of these students is legally entitled to the services specified in their Individualized Education Program (IEP) document, which is co-developed by a child's school team (e.g., teachers, school psychologist), family, and in many cases, the student. The IEP is key for teachers, related service providers, and school leaders as they support students with disabilities. Despite the prevalence of special education, most research on IEPs only focuses on issues of legal compliance or components of IEPs using a small sample (e.g., Kurth et al., 2025; Landmark, 2013). There is emerging research, however, exploring the content of IEPs at scale. Cleveland & Markham (2024) analyzed the content of over 400,000 goals in the digital IEP records of over 150,000 students, finding that goals could be categorized into 10 subjects and 40 skills. Additionally, Cleveland, Kaler, & Markham (2025) used the same sample to explore IEP services, identifying 25 distinct service categories, as well as service descriptors related to proximity (e.g., direct or indirect services), setting (e.g., general education, special education, resource), and group type (e.g., individual, small, large), among others. This work provides a foundation for future analyses using similar records to further explore IEPs at scale. Purpose: This study builds on work by Cleveland & Markham (2024) and Cleveland, Kaler, & Markham (2025), using digital IEP records to classify a statewide sample of IEPs, focusing on services. Our findings have implications for numerous aspects of policy and practice, including but not limited to the challenges that many schools face in adequately staffing special education positions (e.g., Gilmour et al., 2024; Nguyen et al., 2024; Theobald et al., 2025). This study focuses on the service times specified in students' IEPs, which has direct implications for the delivery of special education services and the personnel required to deliver them. Research Questions: (1) What cycles, frequencies, and durations of time are specified in IEP services? (2) How do service times vary based on service category and disability classification? (3) How do service times vary based on district- and school-level staffing? Data: For this study, we use statewide administrative data from Indiana acquired through a research partnership with the state department of education. We analyze comprehensive text data from the IEPs of all students in the state in the 2022-2023 school year. We extract text data that specifies the services included in each IEP, totaling 459,703 unique services. We link IEP data with individual-level administrative data on students' demographics (e.g., race/ethnicity, gender), primary and secondary disability classifications, and grades. Additionally, we link this data to school- and district-level administrative data on personnel (e.g., teachers, paraeducators, social workers, bus drivers). Method: In our analysis, we classify each unique service using a series of text-based descriptors, including the type of service that a student receives (e.g., counseling, speech, transportation), which we term "service category." Each service has an associated cycle, frequency, and duration. Below is an example: [X] will receive direct instructional supports with his math problem solving skills via push in the general education setting and/or pull out to a separate setting for small group/one-on-one assistance. In the example above, the cycle is "per week," frequency is "2 times," and duration is 60 minutes per week. We aggregate all durations to annual hours -- 42 hours per year in this example -- to analyze services of different cycles and frequencies across a shared metric. To analyze the relationship between annual service hours and disability classification, we use a linear regression model of the following form: (1) Y[subscript ig] = B[subscript 0] + B[subscript 1]Disability[subscript i] + X[subscript i] + [lower case final sigma][subscript g] + [epsilon][subscript ig]. Y is annual hours of service time, and "Disability" is a categorical variable for primary disability classification. X[subscript i] is a vector of demographic characteristics, including secondary disability, race/ethnicity, gender/sex, English learner status, and poverty status. [lower case final sigma][subscript g] is a fixed effect for student grade. This model uses specific learning disability (SLD) and pre-kindergarten as the baseline categories. Standard errors are clustered on the student's school. We also explore how annual service time varies by service category using the following model: (2) Y[subscript ig] = B[subscript 0] + B[subscript 1]ServiceCategory[subscript i] + X[subscript i] + [lower case final sigma][subscript g] + [epsilon][subscript ig]. In the model above, Y is annual hours of service time and "ServiceCategory" is a categorical variable for service category. This model uses an aggregate category of "all other services" and pre-kindergarten as the baseline categories to examine the relationship between service time and the six most high-frequency services in the sample (i.e., specially designed instruction, speech-language pathology, occupational therapy, transportation, physical therapy, and counseling services). Covariates, fixed effects, and clustering of standard errors in model 2 mirror those of model 1. Finally, we use a similar linear regression model to estimate the relationship between annual service hours and district- and school-level staffing allocations, accounting for disability type. Findings: Our analysis indicates that students with SLD receive about 100 hours per year on average. Students with intellectual, emotional, and multiple disabilities receive the most service hours annually, receiving about 50-150 hours more per year than students with SLD. Students receive the most annual hours for services specified as occurring on a daily cycle, rather than weekly, monthly, or per reporting period. Conclusions: These findings indicate that there is variation in the ways that schools organize to implement special educational services. For example, enrollment variation for different disability classifications and the determination of service delivery influence the FTE and type of personnel needed to deliver service hours. Overall, this work builds upon Cleveland & Markham (2024) and Cleveland, Kaler, & Markham (2025) by elucidating the nature of IEP services assigned to students with disabilities. This is essential information for policymakers and practitioners, as IEPs hold key information about special education for students with disabilities on an individualized level. As this study demonstrates, analyzing such data on an aggregate, statewide level can bring important insights into how special education services are allocated to students with disabilities and how schools may best organize themselves to deliver services.