Clery Act

The Jeanne Clery Disclosure of Campus Security Policy and Campus Crime Statistics Act (20 U.S.C. § 1092) was signed in 1990 after Jeanne Clery was raped and murdered on the Lehigh University campus. The Act mandated disclosure of campus crime statistics and was built on a philosophy that, had Jeanne Clery and her parents known the LU crime statistics, she would have gone to a different school. As stated in a program review by the Department of Education: the Clery Act is “based on the premise that students and employees are entitled to accurate and honest information about the realities of crime and other threats to their personal safety and the security of their property (US Department of Education).” This legislation reflects a particular political moment as does it embed ideas about crime rates and crime prevention, etc.

Over the last 30+ years, little empirical data has been gathered regarding the effectiveness of the Act. One particular area of interest to our project is the student perceptions of the emergency notifications that UC Berkeley sends out in part in order to comply with a 2007 update to the Clery Act which was motivated by the Virginia Tech shooting. Additionally, in September 2020, UC Berkeley was fined by the Department of Education 1.35 million for violating Clery including “failure to issue emergency notifications” and “failure to issue timely warnings in accordance with federal regulations” (US Department of Education).

All information regarding the Clery Act has been compiled by UC Berkeley students for members of the UC Berkeley community and those wishing to understand and analyze how Clery is enacted within their communities in hopes that the more we understand how and why these systems succeed and fail the more we can improve them and keep our community safe.

Clery Geography

Each public university within the scope of the Clery Act has 4 categories of geography: on-campus, on-campus student housing, off campus, and public property. When we spoke to Abigail Ogden, Director of Clery Compliance at UC Berkeley, she noted the fuzzy boundaries on the map and that notifications will often be sent for incidents within and also just beyond the boundaries of Clery Geography.

Furthermore, there are some interesting implications for coding of various campus-adjacent institutions as within or outside of Clery Geography. For example, if a certain fraternity ceases to be a recognized student organization due to disciplinary measures, it also ceases to be within Clery Geography.

Clery Crimes

The Clery Act requires universities to report statistics annually on crimes that are violent crimes, Violence Against Women Act (2013) offenses, hate crimes, and liquor, drug, and weapon arrests and referrals. What is unclear is the relationship between reported or suspected crimes and those found to be actual crimes within a court of law.

Coding WarnMes

The project collected 2 years of WarnMe messages via public records requests and 4 annotators coded the messages along the lines of location granularity, scene descriptions, and personal identifiers used within the context of victims and suspects. We then calculated Fleiss' kappa, a statistical measure for assessing the reliability of agreement between multiple annotators, selected the highest agreement codes for each WarnMe, and analyzed the frequency of the various codes.

The inter-annotator agreement (a weighted average of Fleiss’ kappa) across the location, description, and personal identifier codes was 0.4828. The range of IAA values was 0.4472 for location to 0.6055 for subject’s personal identifier. This range and the average across all the categories shows moderate agreement between the annotators as Fleiss’ kappa spans from -1 to 1.

Additionally, we performed some exploratory analysis on the WarnMes using the highest agreement codes for each message. One interesting finding was that about half of the messages were coded as referring to the location-based subject of the message as something other than a specific address or named place.

The lack of location specificity in many of the messages speaks to the need for increased transparency and accuracy of reporting in these messages.

You can find the code for this project on GitHub.