How data gaps feed inequity
Fully representative datasets are crucial for creating accurate narratives and targeted interventions.
As state and local agencies invest time and money to create data-driven government, incomplete data may set solutions up for failure.
For instance, inequitable distribution of data collection efforts will produce inaccurate data such as when a city deploys air quality sensors in limited areas. The devices could inaccurately identify the air quality as healthier than it really is because only parts of the environment were assessed, said Gillian Diebold, policy analyst at the Center for Data Innovation.
“When groups are left out—or overcompensated for—in datasets, that’s going to create false narratives,” she said. As a result, efforts to improve the air quality may not factor in a community’s complete needs because some places were unaccounted for.
Fully representative datasets are crucial for targeted interventions and ensuring everyone “correctly” benefits from them, Diebold said.
Disaggregating data can also help government create accurate intelligence. For example, an area may have a high crime rate and therefore be seen as a dangerous neighborhood. However, a breakdown of the data may reveal that the majority of the incidents are petty thefts, not necessarily violent crimes, Diebold said. Disaggregation is especially important for geospatial data, where contextual data layers such as types of crime can give users a more holistic view.
Open data is another way state and local governments can squelch inaccuracies and enhance transparency. Publicly accessible open data portals make it easier for residents to contest false or misleading information if data does not reflect their lived experiences, Diebold said.
Governments should clearly define and label their datasets to avoid reporting inconsistencies. For example, federal agencies often have varying definitions of what constitutes a sexual violence offense, Diebold said. Those differences can create disparities in the number of recorded assault cases and make it difficult to understand the scope of the issue, which could lead to insufficient resource allocation and program evaluation.
If data on sexual assault is incomplete or ambiguous, for example, “a policymaker … can pick a number that they find convenient and say, ‘It doesn’t need that many resources. It’s not that big of an issue,’” she said.
To improve data quality and outcomes, agencies should practice better documentation, such as defining what information is included in a dataset and if it has any limitations, Diebold said. Performing regular audits and evaluations can also reveal where data collection falls short so data stewards can focus on improving those areas.
Inequitable or ineffective data collection “can really create critical data gaps and tell an incomplete or inadequate story about a given issue,” Diebold said.