Complex human behaviors related to crime require multiple sources of
information to understand them. Social Media is a place where people share
opinions and news. This allows events in the physical world like crimes to be
reflected on Social Media. In this paper we study crimes from the perspective
of Social Media, specifically car theft and Twitter. We use data of car theft
reports from Twitter and car insurance companies in Chile to perform a temporal
analysis. We found that there is an increasing correlation in recent years
between the number of car theft reports in Twitter and data collected from
insurance companies. We performed yearly, monthly, daily and hourly analyses.
Though Twitter is an unstructured source and very noisy, it allows you to
estimate the volume of thefts that are reported by the insurers. We
experimented with a Moving Average to predict the tendency in the number of car
theft reported to insurances using Twitter data and found that one month is the
best time window for prediction.