However, several researchers have pointed to methodological problems associated with this research literature. First, researchers have argued that the British car park studies included in the meta analysis cannot accurately control for the fact that CCTV was introduced simultaneously with a range of other security related measures. Second, some have noted that, in many of the studies, there may be issues with selection bias since the introduction of CCTV was potentially endogenous to previous crime trends. In particular, the estimated effects may be biased if CCTV is introduced in response to crime trends. It has been argued that problems of selection bias and endogeneity can be addressed by stronger research designs such as randomized controlled trials and natural experiments. A 2017 review published in Journal of Scandinavian Studies in Criminology and Crime Prevention compiles seven studies that use such research designs. The studies included in the review found that CCTV reduced crime by 24 28% in public streets and urban subway stations. It also found that CCTV could decrease unruly behaviour in football stadiums and theft in supermarkets/mass merchant stores. However, there was no evidence of CCTV having desirable effects in parking facilities or suburban subway stations. Furthermore, the review indicates that CCTV is more effective in preventing property crimes than in violent crimes. Another question in the effectiveness of CCTV for policing is around uptime of the system; in 2013 City of Philadelphia Auditor found that the $15M system was operational only 32% of the time.