Falling standards of journalism and intentional divorce from fact checking has called for evolution of fact based journalism. Data based journalism can be considered to be an attempt towards that path.
Recently 3 events which caught attention of nation viz, Dadari incident, release of religion based demographic data and now increasing crime rate in Bihar were largely based on popular/media perception. (Consulting) journalist Anand Ranganathan pulled out statistics from various sources and refuted popular claims in all three incidents.
While his attempt must be appreciated for setting an example of verifying popular claims with actual statistics, I believe this approach is incomplete and may create a ‘statistical perception’ which is statistics which may not quite match with day to day experiences.
As with any empirical phenomenon data based journalism has fear of under-fitting or over-fitting given data. Under-fitting meaning treating each case as separate and observing no patterns, while over-fitting means drawing tight inferences based on given data which may not fit additional data.
The approach taken by Anand, I think, over-fits his data. In case of Dadri he compared number of communal incidents before and after new government, for religious demographics he took growth rate of different communities, and for Bihar again he took crime stats under 2 governments.
But I believe that this is too simplistic approach where only one dimension of data is considered to analyze emotion. For example, any murder will show up in statistics as addition of just one data point. But impact of a murder on society especially those living in same locality depends on number of other factors. Time of murder, number of witness, type of crime scene (residential, commercial, farm), population density of crime scene, background of perpetrator and victim are few such factors. So, a crime committed in broad daylight in residential area will have way more impact on people’s psyche than death of a crook in gang war outside the city at midnight. Although in crime records, both will carry same weight.
Similarly, in case of demographic data, population of one community will not overtake other in 100 years. So, Anand says that there no room for fear-mongering. But that does not take into account degree of integration of two communities, whether settlements of a community are closely knitted where persons from other community can not think of settling, does change in population has any covariance with crime stats. These are the factors that can have more impact on psychology than surname of person living in next apartment.
So, while any news report must be backed with proper research and data, dismissing popular perception without taking impact factor of different incidents into account is doing disservice to statistical inferences.