Backdrop: This piece is based on a certain incident when a Bangladeshi Lawmaker laughed on national television at the death of 3 children in a road accident, saying that India’s death rate per hour is 16 – and that this is not an issue for us to fuss about.

–

Throwing around controversial statistics like cake is actually a really wrong analytical move – especially more so when policymakers use it.

For instance, yes it is true that 16 people die every hour in India in a road accident. But it is also true that in terms of land mass, India is 22 times the size of Bangladesh. Which means that every hour, hypothetically drawing tangents, Bangladesh’s death rate should hover around 0.727 because we obviously have 22 times the less amount of roadways. In a day, that would be 17.45 deaths. Every year, that is 6370 “acceptable” deaths.

But here is the problem – if you do the same math, but this time considering the population as your pivot instead – your figures change. For a population of 1.342 billion, our Minister’s acceptable death rate is 16 per hour. In that case, hypothetically for our population of 163 million – the death rate should be 1.97 deaths per hour, or around 47 deaths per day. This would mean that 17255 deaths per year is an acceptable value.

The difference between the two numbers is remarkably different. The reason why I say this though, is because it instantly brings to light how this specific set of information is very inadequate when it comes to quantitative analysis, and would hypothetically mislead policymakers and their mental visualization of an issue and its impact on the subsequent policy-making and prevention measures.

You see, while India’s number of 16 per hour sounds extremely far-fetched from our perspective, the real life value is actually in no way different on the ground. According to BJKS, a total of 2123 people were killed in the last 4 months in Bangladesh in road accidents. Every year, that takes the tangent close to 6369 – a number that is star-kingly close to the initial area-weighted estimate of 6370 deaths per year – which hypothetically means that 16 people die in Bangladesh every hour too.

The first time I realized the problem with such rampant throwing-around of numbers was when I was reading a paper on the improvement that the invention of Hilton’s Capsule Network has made over its next best Convolutional Neural Network based alternative. The exact number was that the Capsule Network made an improvement of close to 45% – which instantly makes it look like the Capsule Network made a striking change that would render CNN’s useless over the coming years. The reality behind the number was that a CNN made a mistake 2.5 times out of every hundred, while Hilton’s Capsules made a mistake 1.4 times out of every 100 tries. Which means that they are both still really great options – unless you consider the small intrinsic gains.

—

So yeah, before you throw around statistics like that – please understand the relative weight behind your nodes. Or else you will just sound stupid on national television.