**Principal investigator:** **Gary Kleck** Florida State University Email: gkleck@fsu.edu Homepage: http://www.criminology.fsu.edu/p/faculty-gary-kleck.php **Sample size:** 5550 **Field period:** 9/5/2009-12/28/2009 **Abstract** A national Internet survey of a probability sample of 5,550 U.S. adults was used to explore some of the sources of error in survey estimates of the frequency of defensive gun use (DGU) by crime victims, using experimental methods. Respondents were randomly exposed to variant question wordings and combinations of questions to judge whether these variations influenced the willingness of Rs to report victimization experiences and DGUs. These findings were obtained. (1) Rs were 70% more likely (significantly so) to report a victimization experience when they were explicitly instructed to report incident involving offenders known to them. Since prior DGU surveys have not provided such instructions, it is likely that they missed DGUs against offenders known to the R. (2) Rs were 43% more likely (significantly so) to report a victimization if they were instructed to include incidents that turned out alright for the R, i.e. resulted in no injury or property loss to the victim. Since previous DGU surveys did not include such instruction, this suggests that they were especially likely to miss successful DGUs. (3) Rs were no more likely to report a DGU for the previous five years than for the previous 12 months, even though there should have been far more reports for the longer recall period. This suggests that long recall periods produce substantial underreporting of DGUs, due to recall failure. (4) Rs were 125% more likely (significantly so) to report DGUs if they were directly asked about DGU than if they were first asked about victimization experiences, and then asked about DGU in connection with reported victimization experiences. (5) There was no significant difference in reporting of DGU depending on whether the R was asked a general open-ended question about any kind of self-protection action, vs. a question asking specifically about DGU. (6) Finally, there may be substantial effects of survey mode on reporting of DGUs. Although the comparison is not based on random assignment of conditions, the Internet survey yielded a one-year prevalence of DGU experience among U.S. adults of 6.4%, compared to just 1.3% in a national telephone survey done in 1993. **Hypotheses** 1. Rs are more likely to report a crime victimization experience if they are explicitly instructed to report incident involving offenders known to them. 2. Rs are more likely to report a victimization if they are instructed to include incidents that turned out alright for the R, i.e. resulted in no injury or property loss to the victim. 3. In the absence of recall failure, Rs should be far more likely to report a DGU for the previous five years than for the previous 12 months. 4. Rs are more likely to report DGUs if they are directly asked about DGU than if they were first asked about victimization experiences, and then asked about DGU in connection with reported victimization experiences. 5. Rs are more likely to report a DGU if they are asked a question asking specifically about DGU than if they are asked a general open-ended question about any kind of self-protection action. **Experimental Manipulations** Alternate wordings of questions and presence/absence of accompanying instructions were randomly varied. The specific conditions varied are given in the Hypothesis/Research section. **Outcomes** Reporting of a defensive gun use experience. **Summary of Results** 1. Rs were 70% more likely (significantly so) to report a victimization experience when they were explicitly instructed to report incident involving offenders known to them. Since prior DGU surveys have not provided such instructions, it is likely that they missed DGUs against offenders known to the R. 2. Rs were 43% more likely (significantly so) to report a victimization if they were instructed to include incidents that turned out alright for the R, i.e. resulted in no injury or property loss to the victim. Since previous DGU surveys did not include such an instruction, this suggests that they were especially likely to miss successful DGUs. 3. Rs were no more likely to report a DGU for the previous five years than for the previous 12 months, even though there should have been far more reports for the longer recall period. This suggests that long recall periods produce substantial underreporting of DGUs, due to recall failure. 4. Rs were 125% more likely (significantly so) to report DGUs if they were directly asked about DGU than if they were first asked about victimization experiences, and then asked about DGU in connection with reported victimization experiences. 5. There was no significant difference in reporting of DGU depending on whether the R was asked a general open-ended question about any kind of self-protection action, vs. a question asking specifically about DGU. 6. Finally, there may be substantial effects of survey mode on reporting of DGUs. Although the comparison is not based on random assignment of conditions, the Internet survey yielded a one-year prevalence of DGU experience among U.S. adults of 6.4%, compared to just 1.3% in a national telephone survey done in 1993. **References** Kleck, Gary. 2010. "How Prevalent is Defensive Gun Use? Results from a National Internet Survey Experiment." Paper presented at the annual meetings of the American Society of Criminology, San Francisco, CA, November 19, 2010.