Why this project, and what it can honestly show.
California publishes elder-protection data, but it does not say what that data can support. This page sets out why the Observatory exists, the questions guiding it, the research that shaped its focus, the outcomes it hopes to encourage, and why a data-science approach was necessary.
Why this project
Elder financial exploitation is deeply harmful, and it tends to find the people least able to absorb the loss: older adults who are isolated, unwell, or dependent on someone else for care. I started this project because I wanted to know what California's public data could actually tell us about that harm, rather than what we might assume about it.
The Observatory therefore works through a small set of questions. How do reported elder financial-abuse allegations change within counties over time? How far does small-cell suppression limit what can be learned from public Adult Protective Services records? And which patterns can be interpreted responsibly, given what the published data does and does not contain?
Answering those questions required a data-science approach: combining monthly administrative records with population denominators and suppression-aware statistical methods, applied consistently across all 58 counties.
The work is meant to be useful beyond the analysis itself. It is offered as a contribution to APS data transparency and to more careful public reporting about what county-level elder-protection statistics can and cannot support.
Research Framing
The project is guided by one central question: can publicly available SOC 242 data support reliable county-level analysis when small-cell suppression, population differences, and variation in local reporting systems are explicitly considered? Within that, it asks how reported elder financial-abuse allegations change within counties over time, how suppression is distributed across counties, and how sensitive county-level conclusions are to the way withheld small counts are handled.
The starting point is the prevalence literature. A systematic review and meta-analysis by Burnes and colleagues (2017, American Journal of Public Health) estimated the scale of financial fraud and scams among older adults in the United States, and the New York State elder abuse prevalence study, Under the Radar (Lifespan of Greater Rochester, Weill Cornell Medical Center, and the NYC Department for the Aging, 2011), showed that elder mistreatment is substantially underreported, with only a fraction of incidents ever reaching a formal system. A 2025 systematic review by Mosafer and colleagues (Geriatric Nursing) further documents the demographic, health, and social factors associated with financial exploitation. Together, this work is why the Observatory treats APS records as reported allegations rather than estimates of true prevalence, a distinction the U.S. Government Accountability Office has also drawn in calling for stronger state reporting on the costs of financial exploitation (GAO-21-90, 2020).
A second strand concerns the administrative data itself. Mosqueda and colleagues (2016, Journal of Evidence-Informed Social Work) documented county-to-county variability in how California APS investigations are conducted and how their outcomes are interpreted, which means some of the variation in published statistics reflects agency practice rather than underlying harm. Matthews, Harel, and Aseltine (2017, Journal of Public Health Management and Practice) reviewed statistical disclosure-control techniques in web-based public-health data systems, establishing that suppressed small counts are a known, structural limitation of this kind of published data.
The focus on suppression grew directly out of these two strands: if withheld counts are concentrated in smaller, more rural counties, then direct county comparisons may reflect differences in data availability as much as differences in reported allegations. Two further bodies of work shaped the interpretation. Liu and colleagues (2022, Journal of Family Violence) showed through San Francisco APS administrative data that the COVID-19 pandemic disrupted elder mistreatment reporting, which is why the study period's trends are not read as pure changes in underlying abuse. Spoer and colleagues (2020, Preventing Chronic Disease) demonstrated that county-level health metrics can behave differently from finer-grained measures, a reminder that county-level associations here are ecological and not individual-level or causal evidence. The multiple-testing correction follows Benjamini and Hochberg (1995).
The intended contribution is not a ranking of counties but a suppression-aware framework that distinguishes robust longitudinal reporting trends from conclusions that depend heavily on how withheld counts are handled. For policymakers and administrators, the findings point toward concrete improvements: more consistent county reporting practices, clearer public documentation of how small cells are withheld, and publication formats that let researchers analyze trends without guessing at suppressed values. The Observatory also models what careful public reporting on elder-protection data can look like.
The questions above could not be answered from any single source. They required joining monthly administrative records to population denominators, characterizing suppression patterns across all 58 counties, and applying interval-censored statistical models that treat suppressed counts as ranges rather than invented exact values, with multiple-testing correction and bootstrap uncertainty estimation to keep conclusions honest. A data-science approach was the only way to do that consistently and reproducibly across six fiscal years of records.
How the monthly records, population denominators, and suppression-aware models are combined, and what the results do and do not license.
What six fiscal years of reported allegations show across all 58 counties, stated alongside the caution each finding requires.
The Observatory's mission, the datasets and agencies behind it, and the people and institutions it acknowledges.