This study investigates which crash characteristics influence the probability that the crash is reported in the television news. To this purpose, all news items from the period 2006-2012 about traffic crashes from the prime time news of two Belgian television channels are linked to the official injury crash database. Logistic regression models are built for the database of all injury crashes and for the subset of fatal crashes to identify crash characteristics that correlate with a lower or higher probability of being reported in the news. A number of significant biases in terms of crash severity, time, place, types of involved road users and victims' personal characteristics are found in the media reporting of crashes. More severe crashes are reported in the media more easily than less severe crashes. Significant fluctuations in media reporting probability through time are found in terms of the year and month in which the crash took place. Crashes during week days are generally less reported in the news. The geographical area (province) in which the crash takes place also has a significant impact on the probability of being reported in the news. Crashes on motorways are significantly more represented in the news. Regarding the age of the involved victims, a clear trend of higher media reporting rates of crashes involving young victims or young fatalities is observed. Crashes involving female fatalities are also more frequently reported in the news. Furthermore, crashes involving a bus have a significantly higher probability of being reported in the news, while crashes involving a motorcycle have a significantly lower probability. Some models also indicate a lower reporting rate of crashes involving a moped, and a higher reporting rate of crashes involving heavy goods vehicles. These biases in media reporting can create skewed perceptions in the general public about the prevalence of traffic crashes and eventually may influence people's behaviour.