We are investigating potential biases and external factors in how crime data is collected, focusing on the relationship between sex and crime. In particular, we analyze the dataset’s crime intensity ranking—where lower numbers indicate more violent crimes—and the gender of the victims to question the validity of this categorization system.
What does this data tell us about the relationship between sex and frequency/intensity of criminal activity? We began with a the hypothesis that the portion of crimes committed against women will involve more frequent and violent crimes.
Using Koopman’s anatomies of system-level format architectures, we analyzed what the dataset reveals on a macro, systemic level. This involves questioning the assumptions behind how the data is collected (e.g: how crime severity is defined, how reliable crime reports are, and what identity-based assumptions may shape the data). Our deconstructive analysis suggests possible discrepancies in how crime severity is categorized, raising broader concerns about how similar systemic issues could affect other large-scale datasets.
After deconstructing the dataset, we identified several gaps and inconsistencies. Notably, the perpetrator’s sex is missing, which limits sociological analysis of crime patterns. We also found issues in how crime severity is classified: although lower crime codes (100–200) typically represent more serious crimes, some cases appear mis-categorized; for example, child abuse was coded as 627, while a stolen bike was coded as 480. These inconsistencies raise questions about how severity is determined; since crime codes originate from penal codes created by the California State Legislature, potential reform may depend on the officials Californians elect to legislative positions.
Our sample partially supported our hypothesis: while women were more frequently victims overall, violent crimes were more likely to involve male victims. However, the data only comes from 2022, which may be an unusual, post-pandemic year. Also, past reporting issues, such as a 2014 investigation that found about 1,200 violent crimes incorrectly recorded as nonviolent, suggest possible inaccuracies.
This project focuses on LAPD Crime Data from 2020 to the present. This dataset can be described in general terms as reported crime and arrest data, including place, crime, victim’s details, etc. Our set included about 1,000,000 rows of data, so narrowed it down to a randomized group of 250 cases, which came from the year 2022. We are analyzing this data under Poirier’s deconstructive method, and Koopman’s macro-level formatting.
