That’s what I conclude in a piece I’ve written for Mint which got published today.
I analyze the last digit of the vote tally of different contestants in different provinces, and find an unusually large number of numbers that end in zero – the odds of this happening at random are at most 2.25%, I conclude.
You might be aware that I’ve been doing this series on elections for Mint for over a year now. Since the Indian elections are over, and steam is yet to pick up for state elections in Maharashtra, Haryana and Jharkhand I’m dabbling a bit in analysis of international elections. If there’s something potentially interesting that you want me to analyze, do drop a note here.
Does Afghan constituency sizes have a large range of values? If so, shouldn’t we expect more zeros (a la Benford’s law)?
benford’s law is to do with the first digit, and nothing to do wtih the last. I don’t see why even if it has a large range of values (it doesn’t) 0 should be more probable!
Ah ok – big endian/little endian 😀 – serves me right for not reading this carefully enough.
One more doubt then – my take on the plot is that the bump at 5 is not too significant, so this is mainly re. the bump at 0. Now if I understand, this is data for each constituency/participant? If so, could it be that some participants were super unpopular, and got zero votes in many constituencies, which biases the distribution? Do the results hold up when you set a minimum threshold (say 10), and drop all numbers below it? Alternately, why not just compare the winner and the second place?
Immediately after I sent in my piece, I realized that I hadn’t accounted for zeros! And quickly checked. No, there are no candidates who have received less than 10 votes in a province. So taken care of that particular thing!
We can compare just first and second but that leads to much fewer data points and all that!