In defense of anecdotes

What prompts this post is the discussion taking place at Yazad Jal's AnarCapLib. This would have been ideally a comment over there, but I thought perhaps this one is a bit too long for that. Do check out Yazad's post, and the comments over there.

Yazad's post starts an interesting discussion on whether or not the Indian economic reforms have benefited the poor. In particular did it help in bringing down poverty levels? The post follows Dilip D'Souza's and Yazad's Rediff columns on the same topic. An interesting question that was asked was to what extent one can trust anecdotal evidences.

I like to think that anecdotal evidences do not count. I also like to think that only figures, facts, graphs, and charts count. Then one realizes that, in real life, anecdotal evidences do matter. Many a time only anecdotal evidences matter. I found this out once again recently, when I had to write a reco for a former student. The program chair made it clear that she would like to see my "anecdotes" about the student's overall performance, and I shouldn't be stressing on the student's grades etc. Incidentally, the student was applying for an MS in Statistics! Life would be certainly much easier if everything can be put into percentages and neatly columned tables. But it's too complex for the statistical tools to tame it.

Perhaps that's one reason why having certain figures at hand do not end a debate on social issues like poverty eradication. Obviously, a sizeable population of "otherwise sane" people are not willing to buy figures of this kind in order to come to a conclusion. We do not debate election results here in India because we trust absolute headcount. The very fact that we do debate data on poverty does say that an absolute trust in such data is lacking even from the perspective of some one who endorses the data at hand.

Not to say that statistics do not tell you anything right. They do, but not always. Chances are that a majority of those who are just above that "poverty line" wouldn't buy that statistics! Also which one to buy? For instance, Narasimha Rao govt had maintained two different numbers at the same time; one was twice the other! In another instance, nine consecutive surveys do not show any reduction in poverty, but the tenth one shows a ten percentage points reduction. I believe that statistics can give you meaningful, truth-approximate results, if you care for it. I do not believe in this sort of figures in general as it's almost trivial to manipulate the starting criteria. Must say that Disraeli is often vindicated these days!

So have the reforms benefited the poor? I guess there's no blanket Yes or No answer here. Have the reforms benefited the rich, the upper middle class? You know the answer, right? Now does that say anything? Also, has the inequality increased? Or is it that equality in itself isn't a great thing to have?

Apparently, the voters do not like the reformists. Natural guess is that this is because the reforms haven't benefited the majority. The only way to sell the reforms to more and more people is to tell them that these reforms promise a bright future. But then who doesn't promise a bright future? People wouldn't buy that therefore. Do the govt figures come handy then? Perhaps not. India did shine in figures. That helped India to shine in opinion polls, but she did not, when it came to actual voting machines. Mass propaganda statistics got reflected in mass media statistics, but beyond that, statistics did not count. Perhaps in elections too, as in the case of the reco I wrote, it's anecdotal evidences that matter.