In 2015, we created the Incarceration Trends data tool to fill a critical gap: the lack of a comprehensive resource with historical jails data for every U.S. county.

It was around this time that Vera decided to focus on jails as a way of addressing mass incarceration. The creation of this database was a critical means to both problem spot runaway jail population growth—such as in Indiana—as well as broader trends, such as the shifting geography of incarceration from urban to rural places and the disproportionate growth of women in jail.

The Incarceration Trends data is sourced from surveys that local jails have been completing for the federal government for more than 45 years. The development of the database was challenging because the survey data was never designed to be “stitched together.” Over time countless jails have opened and closed, some of the questions on the survey have changed slightly, and even county boundaries have changed.

Because of the structural challenges of working this this data, we are continuously learning and making improvements to the tool. Last week, for example, we learned of a very uncommon problem with the data. During a “town hall” event on CNN, presidential candidate Sen. Amy Klobuchar (D-MN) claimed that the African-American incarceration rate dropped 65 percent in Hennepin County during her tenure as the county’s lead prosecutor from 1999 to 2007. Following the interview, the Senator’s campaign pointed to the jails data on our Incarceration Trends data tool as the source. Subsequent fact-checking by Politifact and The Washington Post noticed it was conspicuous that the dramatic decline in African-American incarceration occurred in just one year and that the county’s jails population, when disaggregated by race, summed to less than half the total jail population.

There was indeed a problem with the data. When we investigated, we found that, in 2001, Hennepin County stopped reporting the race of anyone held in one of the two jails in the county. One jail, however, did continue to report data on race. Because we sum the data for all jails in a county, the African-American jail population artificially declined.

This problem with the data is very uncommon. Only about 2 percent of jails do not report race data, and because the vast majority of counties only have one jail, the data is usually just missing in these instances. But because there are two jails in Hennepin County, one that reported race data and one that did not, we displayed partial data after 2001 thus creating the illusion of a decline.

We are in the process of making an improvement to the website that will address this problem, and will let users see when race data is partially or completely not known for a county. For Hennepin County, for instance, this would indicate that since 2001, more than half of the jail population was missing race data.

Jail data is notoriously difficult to work with. Thousands of jails systems across the country use a variety of data management systems. And while conversations about jails often focus on pretrial detention, the reality is that jails are used differently across the country. Some are nearly exclusively used to hold pretrial detainees, while in others the majority of the people held are serving a sentence. Adding further complexity, some jails rent beds to other state, local, and federal corrections systems—thereby influencing the county’s growth trends and incarceration rate on a per capita basis. Furthermore, when people are transported from one jail jurisdiction to another, they are “counted” in the jail where they are held, thereby undercounting the number of people detained by county justice systems that use other counties’ jails.

Consequently, it is usually valuable to review a county’s prison admission and prison population data—also available in the Incarceration Trends data tool—to get a “reality check” of the jails data. Because nearly everyone who goes to prison passes through a local jail first, prison trends are often similar to jail trends, and dramatic differences merit further inquiry. Prison data, which is sourced from the administrative records of state prison systems, is generally more complete and comprehensive than the jail survey data.

In the coming year, Vera will continue to work with state and local governments and advocates to identify gaps in data, use publicly available data sources to share fresher insights, and innovate and expand data collection across the country. We believe that transparent, accurate, timely data about how our nation is using local jails and the way in which they’re impacting communities is vital—for a healthy democracy, to challenge assumptions, and to craft truly meaningful reforms at the state and local level.

But we also recognize that data alone does not tell the whole story. The lived experiences of communities and individuals whose lives have been directly impacted by the rise of mass incarceration are another way to “reality check” what we see in the existing data, and to help guide future policy. So, as we continue to explore the data, we do so in a broad spirit of partnership with civil society and civil servants alike, with the shared goal of using data and knowledge to reverse mass incarceration. If that ambitious goal speaks to you, or if you ever have questions about what you’re seeing, we certainly hope you’ll be in touch. You can always reach us at backyards@vera.org.