Measures and methods

The two sources

Center reports: Administration for Community Living, Centers for Independent Living Program Performance Report public datafiles, report years 2016 to 2023. Every Center for Independent Living that receives Part C funding files one each year. ACL publishes the numeric sections for all centers as a spreadsheet, two or more years after the period ends, with the narrative and the county table removed. The pipeline reads the 8 yearly workbooks, matches fields by their header text (the headers drift year to year), and keeps one row per center per year.
Census denominators: U.S. Census Bureau, American Community Survey 2019-2023 5-year estimates, tables B18101 and B18101A-I (civilian noninstitutionalized population). The table gives, for every state and county, the civilian noninstitutionalized population and how many of them report a disability, by age band. That is the population a center could in principle reach.

The report-year offset

A CIL PPR labelled report year N covers 30 September of year N to 29 September of year N+1. Report year 2023 therefore ends in September 2024. State (ILS) reports labelled year N follow the ordinary federal fiscal year, so a CIL report year N lines up with state year N+1. When a council compares a center's own report with the state's report for "the same year", it must shift one of them by a year, or the dollars and the consumers will never agree.

Seven families of measure

The Program Performance Report was not built to measure effectiveness. It counts inputs, activity and self-declared goal attainment. It has no comparison group, no follow-up and no verification. What a national panel allows is relative measurement: a center against 348 others, and against its own history. Each family below is one view of that.

Reach: Consumers served as a share of the population with a disability, shown per 10,000. Also information and referral contacts, kept separate because a center can be busy on the phone while serving very few people intensively. Reach is computed for states, because the public file has no county detail for centers.
Intensity and cost: Total income per consumer and per goal achieved. The report gives income, not spending, so these are stand-ins. A low figure is not automatically good: a center helping people leave nursing homes should cost more per person than one that mostly answers questions.
Outcome: Goals achieved as a share of goals set; records closed with all goals met as a share of all closures (recorded at exit, so harder to inflate); and relocations from nursing homes and institutions per 100 consumers and per $100,000 of Part C. The transition count is the one outcome with a bright-line meaning.
Service responsiveness: Consumers receiving a service divided by consumers requesting it, by service type and overall, excluding information and referral. Unmet demand by service type is the most policy-relevant number in the instrument.
Equity: Consumers' race and ethnicity against the state's disability population by race (ACS tables B18101A to I), summarised as an index of dissimilarity from 0 (identical) to 100. The comparison is with the whole state, not the center's service area, so it is a prompt for a question, not a verdict.
Compliance and capacity: Board members with significant disabilities (the rule is more than 50%), staff with disabilities, and total staff. This is the one part of the report that maps to a regulatory requirement.
Stability: Year-to-year variation in total income and in Part B (coefficient of variation), and the share of income that is Title VII. Large swings in Part B are usually a finding about state administration, not about the center.

Percentiles and peer groups

A percentile is the share of centers the center ranks above, after turning every measure so that 100 is the favourable end. Two are shown for each measure.

National: among all 349 centers reporting in 2023.
Size peers: among centers in the same quartile of total income. The quartile boundaries in 2023 are $591,341, $1,178,759 and $2,172,432.

The island and non-contiguous jurisdictions (AK, AS, GU, HI, MP, PR, VI) are marked on their pages. For an island network they are the structural comparators; a ranking against the mainland average is not meaningful.

Fewer than five values in a peer group means no percentile is shown.

Data-quality checks

Nobody is required to make a center's tables agree with each other, so the pipeline checks. Eight rules run over every center-year. A failed rule is a flag; the flag count is itself a measure, and it is shown beside every comparison so a reader knows how much weight the other numbers can bear.

Consumer records do not add up. Records carried over plus new records opened should equal consumers served.
More received a service than requested it. For at least one service type, the count receiving exceeds the count requesting.
Goals achieved exceed goals set. For at least one goal category, more goals were achieved than were set.
Age table does not match total served. The age bands do not add up to consumers served.
Race table does not match total served. The race and ethnicity rows do not add up to consumers served.
Disability table does not match total served. The disability-type rows do not add up to consumers served.
Closure reasons do not match total closed. The closure reasons do not add up to the total number of records closed.
Goal achievement at or above 98%. With 20 or more goals set, nearly every goal was achieved, which can mean goals are recorded as set and achieved at the same time.

In report year 2023, 29 of 349 centers (8.3%) failed at least one check. Rates by rule and year.

What is excluded, and why

Report year 2015, because its grant numbers used an older scheme that cannot be joined to later years.
Statewide aggregate reports (grant numbers ending ILST). A few states file one report for the whole network. They are not centers and would distort every comparison.
Centers with no 2023 report. They appear on their state's page as inactive, with the last year they filed.

A center is identified by its grant number without the two-digit year prefix. The state it serves is characters five and six of that grant number. The State column in the federal file is the headquarters state, which is wrong for a center that operates grants in more than one state; filtering by that column loses those centers.

Limitations, in these words

Any product built from these pages should carry these.

The PPR is unaudited self-report. No federal agency verifies the consumer counts.
Definitions of "consumer served" and "goal achieved" are not applied the same way at every center. The data-quality measure is shown beside every comparison for this reason.
Part B and Part C outcomes cannot be separated. Centers report total activity, and no public source shows what Part B specifically funded.
The public files report income, not expenditure. Cost measures use income as a stand-in and should be checked against audited expenditures before they inform a funding decision.
The public files omit county, city and island detail, and the narrative sections. County-level reach needs data requested directly from each center or from ACL.
The latest public year is 2023 (through 29 September 2024). ACL posts each year's file roughly two years after the period ends.
There is no consumer identifier, so a person served by two centers is counted twice. A state total is an upper bound.
Centers that file a single aggregate report for a whole state network (grant numbers ending ILST) are excluded, because they are not comparable to a center.
Report year 2015 is excluded because its grant numbers used an older scheme that cannot be joined to later years.
Effectiveness here means relative performance against peers and against a center's own history. Nothing in the data supports a causal claim.
ACS counts the civilian noninstitutionalized population. People in nursing homes, prisons and other institutions are not in the denominator, so reach is overstated where a network serves many institutionalized people.

The pipeline is open: read the scripts or download the tables.