Although Did Tells and Crossroads is creative by using a terrorist bomb which lead to police deployment and get statistically significant result, one concern here is that a single shock does not generally allow a researcher to use critical aloes from standard asymptotic theory to judge the statistical significance Of observing a given test statistic.
However, Click and Tabor use terror alert level which varies four times in the period, and it reduces the possibility of spurious correlation. Furthermore, they use the daily data and the treatment window is short, so their results are less likely to be due to changes in other factor. Comparing the earlier studies like Levity(2002), Jovanovich and Sloan(2002), Corona and McCann(2005) which use observational study, Click and Tabor SE natural experiment that usually allow firmer causal conclusions to be drawn than observational studies, so they provide better evidence for a general link between police number and crime rates. Click and Tabor also find no deterrence effect for homicide while Levity found the largest deterrence effect for homicides in earlier studies, this is notable. One significant shortcoming of Click and Tabor research is its reliance on single jurisdiction. Although strong designs require this and it can provide a better identification, there is a concern about external validity. For example, Click and Tabor reveals a statistically significant result applying the data from one single city Washington DC, but it is unclear whether this effect observed in Washington DC could be expected in other cities.
In particular, this research does not provide any evident about the actual variation in terms of the police number and it does not examine how police are allocated to reduce crime. Therefore, use of this design in other cities and more detailed information on police allocation has the potential to increase confidence in these results substantially. This research uses IV method, so a possibly excessive faith in instrumental variable can really be a problem.
