Saturday, March 22, 2008

Left-hand statistics

In "Horrendous Journalism" I recently wrote about the absurd misuse of statistics in journalism. Recently Mr Sitaram Yechury has provided an interesting new example in HT, March 20. I usually enjoy Mr Yechury's reasoned and thought-provoking articles, providing an antidote to the cheerfully irresponsible Vir Sanghvi school of journalism (motto: "let them eat caviar"). But, as the great historian (and TIFR mathematician!) Damodar D. Kosambi is alleged to have once said, "Marxism is not a substitute for intelligence". Herewith Shri Yechury:

"globalisation... has been accompanied by growing economic inequalities... Forty percent of the world's population living on less than $2 a day accounts for 5 percent of global income while the richest 20 percent accounts for three-quarters of the world's income."

I have no dispute with the numbers, for which he cites the Human Development Report 2007-08 and which are grim and tragic. The problem is, they don't prove his point that inequalities are growing. By citing the statistics for a single year, how can you possibly show that inequalities are growing?

Possibly recognising this deficiency, he goes on later in the article to provide some numbers for different time periods: "the wealthiest 1 percent of Americans earned 21.2 percent of all income in 2005. This increased from 19 percent in 2004 and exceeded the previous high of 20.8 percent in 2000. In contrast the bottom 50 percent earned 12.8 percent of all income, which was less than 13.4 percent in 2004 and 13 percent in 2000." (I have taken the liberty of correcting a typo, the published article had "2.8 percent" instead of "20.8 percent").

So what to make of this? Let's re-phrase the confusing presentation as a (good) scientist would, in the form of a table:

Rich (% owned)20.81921.2
Poor (% owned)1313.412.8

This shows that in the period 2000-2004 things actually got better. The rich owned less and the poor owned more. In the subsequent year the trend got reversed. Now unless globalisation was officially launched in 2004, this does not prove Mr Yechury's point. More important - from his article were you able to read off that things got better over a 4 year period? Or did the fuming tone and confusing presentation serve to conceal the facts from you?

I do want to emphasise in conclusion that this should not distract us from the overall tragic picture. Is this horribly skewed income distribution - leading to so much suffering - the best that human beings can manage after so many centuries of religion, philosophy, ethics, science and engineering?

1 comment:

Rahul said...

Well, Sunil, I take your point about each person using statistics for their own end -- and Marxists will be Marxists. But leaving aside that issue, given how small the changes are (both for the poor and for the rich) and considering absolutely no error bars are usually supplied with these numbers, I wonder if the change (whether for better or for worse) are at all significant. The changes are less than a couple of percent in all cases and for the poor a barely .4% between 2000 and 2004.

Given that this is presumably based on some small sample meant to represent millions (at least the 'poor' part of the statistic), wouldn't it be reasonable to assume an error of a few percent?

I would probably therefore venture to suggest that this change upwards or downwards is *not* statistically significant. Perhaps the original HD report will give some idea of the errors in the sampling.

Naively, I would say people in India (except in deeply depressed areas) are living better - I don't mean us, but say maids, drivers, cleaners and others we come in contact with, compared to the time I was a child growing up in Delhi where abject poverty was the norm amongst these people -- but this is hardly a substitute
for hard numbers. And yes, the rate even if positive, is extremely slow.