Liars, damned liars and those who twist numbers
Statistics are just numbers. There’s something intrinsic to numbers that call them into question. It’s how they’re used and by whom that raise doubts.
It’s not clear he actually originated it, especially because he himself attributed it to Benjamin Disraeli. But nobody else knows of Disraeli using it, so the phrase is usually attributed to Mark Twain himself.
As he once wrote: “Figures often beguile me, particularly when I have the arranging of them myself; in which case the remark attributed to Disraeli would often apply with justice and force: ‘There are three kinds of lies: lies, damned lies, and statistics.’”
Statistics, Twain or Disraeli or whoever meant to say—and if you need me to spell it out — are the worst sorts of lies. And that reputation has stuck. People feel they can’t trust them. That there’s something intrinsic to numbers that call them into question.
Yet of course, statistics are just that: statistics. Numbers. Symbols on a page. It’s how they are used, by whom they are used, that raise questions. Seen in that light, statistics themselves don’t lie. They can’t. But what of those who use them in ways that deceive? If inadvertently, that’s one thing. Everyone can be wrong sometimes. But if not? Then they lie. Period.
Take the report I found in a newspaper a few years ago. Sadly, I’ve lost the clipping, so you’ll have to take my word for it. Over the previous year, the headline claimed, car sales in India had, for the first time, outstripped two-wheeler sales. Now this seemed impossible at best, given that in India, two-wheeler sales have hovered for years at about 10 times the number of cars sold. How could that suddenly turn on its head, and so dramatically? Yet here was news saying just that had happened.
But of course, it hadn’t happened. On reading the report, I understood. The sales numbers had not turned around— there was still that approximately ten-to-one ratio between them. Instead, it was the growth. That is, the percentage growth in the number of cars sold in India, year-on-year, had for the first time overtaken the percentage growth in the number of two-wheelers sold.
To make this clear, let’s say two-wheeler sales grew from 25 million in 2017 to 29 million in 2018, while car sales went from 2.5 million to 3 million (indulge me: I made up these numbers, only to make this point). You will note then that two-wheeler sales in 2018 (29 million) are still about 10 times car sales (3 million). But two-wheeler sales grew by 16% between 2017 and 2018, whereas car sales grew by 20%. Now you would hardly take this to mean that car sales zoomed past two-wheeler sales.
Yet what if an unthinking reporter took these growth figures, came to an uninformed conclusion and reported that as news? Well, that’s just what happened with the news report I’m speaking of. No malicious intent there, I’m sure. But it’s likely that a certain lack of familiarity with numbers led this journalist to make a totally false claim.
What does this say about “lies, damned lies and statistics”?
Another example came about, I’m just as certain, actually because of a familiarity with numbers. In particular, a familiarity with how to misuse them to fool people.
I’m writing, of course, of two Bharatiya Janata Party (BJP) tweets that caused some little social media fluttering earlier this week. After a deal of anger over rising diesel and petrol prices, the BJP’s official Twitter handle, @BJP4India, decided to tell us by tweet what they called the “Truth of Hike in Petroleum Prices.” One tweet was titled “% Increase in Diesel Prices”; the other was “% Increase in Petrol Prices.”
Both had diagrams: bar graphs with four bars. Three of the bars are green, the fourth is yellow. Left to right, they are labelled “16 May 2004” (green), “16 May 2009” (green), “16 May 2014” (green) and “10 Sept 2018” (yellow). At the bottom is this line: “Retail Selling Price in Delhi”.
From the right, Narendra Modi looks out at us all.
And what are the numbers attached to each bar? In the case of petrol, they are ₹33.71 in May 2004, ₹40.62 in May 2012, ₹71.41 in May 2014 and ₹80.73 in September 2018. This means an increase of 20.5% from 2004 to 2009, 75.8% from 2009 to 2014, and 13% from 2014 to now—all mentioned on the graph.
The point of this graph? That while the price of petrol has indeed risen, it has risen more slowly over the past four years than during the previous 10 years. The price of petrol has been rising for years, as it must, but this BJP-led government has controlled the rise more effectively than the previous Congress-led one. A simple enough message, certainly, and one that makes sense.
Though if you think about it a little, there is a small problem: the previous time intervals—for the 20.5% and 75.8% increases—were each five years, whereas the 13% rise is across a little over four years. Still, if you extrapolate that 13% to five years—to May 2019—you’d get 15%: substantially less than 20.5%, let alone 75.8%. The message still holds.
The diesel bar graph is similar. There are prices on each of the bars, and across the same time intervals, it shows us 42%, 83.7% and then 28% increases. The same message, again.
So far, so good. With my description so far of these two graphs, you’d certainly agree that they are true to the numbers they represent. Unlike the news report I mentioned earlier, they don’t confuse absolute numbers with percentage growth. Yes, so good.
Except, you’ll note I said “so far”. I haven’t finished describing these two graphs. But now, I will. You should look at them for yourself, but I’m having fun of a kind explaining exactly what they depict. Bear with me while I tell you the rest, and forgive me for the detail.
In the petrol bar graph, the three green bars increase in height from left to right, just as you would expect for the increasing prices they stand for—₹33.71, ₹40.62 and ₹71.41. The percentage growth is shown via arrows from the top of each bar to the top of the one to its right—two arrows labelled 20.5% and 75.8%, respectively. Naturally, these arrows point upwards.
But then we come to the yellow bar, representing the price of petrol now, under the BJP rule. Labelled ₹80.73, you’d expect it to be the tallest of the four bars. You’d expect the third arrow to point upwards, just as its green predecessors do. Right?
But no: the yellow bar for ₹80.73 is shorter than the green bar for ₹71.41, lower than the green bar for ₹40.62, and is in fact nearly the same height as the green bar for ₹33.71. Naturally, given this lower height, the arrow representing the 13% increase points … yes, downwards.
The diesel bar graph gets exactly the same treatment.
Consider what has happened here. The BJP has shown us an increase as a decrease. They’ve represented the highest price on the graph by a bar that’s about the same size as the lowest. What’s the impression when you look at these graphs? Unmistakably, that petroleum prices have fallen under BJP rule, and in fact to nearly 2004 levels. The bar graphs convey as much, the downward arrows underline it, and there’s nothing left to your imagination.
The idea of a fall in prices is exactly what the BJP wants you — or at least, its faithful supporters who ask no questions—to take away from these graphs.
George Orwell would know: this effort is right out of his 1984 playbook.
There is only one word that fits such an effort, and that word is “dishonest”.
So in the end, let’s ask again: what does this say about Twain’s “lies, damned lies and statistics”? For once, I disagree with him.
There are numbers, there are statistics and I know I can’t be as pithy as Mark Twain, but truly there are three kinds of liars: liars, damned liars and liars who twist numbers.
Once a computer scientist, Dilip D’Souza now lives in Mumbai and writes for his dinners. His Twitter handle is @DeathEndsFun