Objectives: Our study questions the common assumption of random DK responses in criminology survey data and emphasizes the importance of understanding and handling DK for gaining substantive criminological knowledge. It examines the individual-level and neighborhood-level correlates of the propensity to give the DK response to questions on individual perception of neighborhood crime in Chongqing, China.

Methods: We designed and conducted an original survey of 4839 residents from 100 urban neighborhoods in Chongqing, China in 2016. Random intercept hierarchical linear models were used to examine the effects of individual-level variables on uncertainty towards neighborhood crime and the effects of neighborhood social process variables, controlling for neighborhood composition.

Results: At least in some instances, DK appears to be the most valid response, reflecting actual uncertainty and lack of knowledge about neighborhood crime. DK responses have substantive correlates at both the individual and neighborhood level. Of particular interest, neighborhood social cohesion is negatively associated with individual uncertainty about neighborhood crime, controlling for neighborhood composition. There is a significant interaction between neighborhood semi-public social control and neighborhood poverty in predicting DK.

Conclusions: Understanding the meanings behind DK has important implications for whether to include the DK option in survey designs and how to handle DK responses in data analysis when they occur. When DK is a valid answer for many respondents, not including the DK option in the survey instrument forces respondents to choose a nonexistent answer, which can result in misleading interpretations.