So imagine my surprise the other day when I read this Business Standard article trying to show Rahul Gandhi ahead of PM Modi using Google Trends data. Because I have myself been following Google Trends data on this for a while and they show no such thing.
So I had to check this out. It is a good thing that Business Standard provided a screenshot, which let me see the period they were checking for, which happened to be Jan 1, 2018, to Jan 6, 2019.
Okay, so I checked it out on Google trends for the same period.
Huh? All I see is “Modi” uniformly ahead of “Rahul”. What is Business Standard even talking about?
To be fair, search terms can vary. There are many ways people could be wording their searches for PM Modi and the Congress President Rahul Gandhi. I started with “Modi” and “Rahul” because those are the shortest ways people generally use to talk about these two people.
But we need to “shop around” with some variants before we can confirm a trend. How about the respective full names: “Narendra Modi” and “Rahul Gandhi”? Looking closely at the screenshot that Business Standard provided, they seemed to have tried out the full names.
So here is the data.
Again, Narendra Modi is almost uniformly ahead of Rahul Gandhi. An exception occurs between Dec 9 and Dec 15, 2018, right after exit poll data came out and Congress went on to form governments in 3 states.
Okay, let’s try something else to try and confirm the findings of the Business Standard. Google Trends measures two kinds of searches: these are “term searches” and “topic searches”.
As Google explains, “term searches” can be very imprecise and cover all sorts of related information, which may be irrelevant. A search for “banana sandwich” may lead to “peanut butter sandwich”! If you want more precise searches, you need to look at “Topic searches.”
So I decided to try a precise “Topic search” for Narendra Modi vs Rahul Gandhi. Here is the result.
Ouch! It got really bad for Rahul Gandhi once I tried the precise “topic search”. The PM is not only well in the lead, but his lead is actually increasing.
So what am I doing wrong that I can’t seem to find the “data” that Business Standard was providing? For this, I decided to read the fine print in the article. Here’s what I found.
Ah…so they pared down the data to the news segment only, not across all categories. Okay, let me do that:
Again, PM Modi seems to have a clear and very uniform lead. Rahul Gandhi hasn’t come close all year.
Okay, enough with the precise “topic searches”. Let me go back to the imprecise “term search” data and see what happens.
It does look better for Rahul Gandhi but PM Modi is still in the lead. There were only 3 periods of the year when Rahul’s numbers surpassed those of the PM. One is when the Congress formed its government in 3 states this December. Another has come between May 13 and May 19. We can immediately recognize this as the period when exit polls wrongly predicted a win for the Congress in Karnataka. So at least 2 of these 3 periods can easily be dismissed as momentary surges of interest in Rahul.
But still, we can find no sign of the “data” that Business Standard was offering us. How did they find Rahul to be in the lead? Looking carefully in the screenshot, I managed to decode what they meant by “news segment”. I was looking in the category of “news”; instead, I needed to look at “news search”.
Huh? Modi is still way ahead! What happened here?
Oh, I know! I put “Modi” and “Rahul” again. That’s not what Business Standard did. I have to use full names! So I did.
What? Modi ahead? Again.
Oh, I know! I am so absent-minded that I missed using the exact specifications required to discover a lead for Rahul Gandhi! I used the precise “topic search” data instead of the imprecise “term search”.
So this time before clicking away, I decided to make a full checklist so that I don’t miss something.
(a) I need to put the full names “Narendra Modi” and “Rahul Gandhi”. Can’t use “Modi” and/or “Rahul”
(b) I can’t use the precise “topic searches”. I must use the imprecise “term searches”.
(c) I can’t use web searches or youtube searches. I must restrict to the news.
(d) I can’t go to the category of news. I must use only the “news search” filter.
With this careful list of filters, I was finally able to get Rahul Gandhi to move ahead of Narendra Modi in Google Trends.
Congratulations to Rahul Gandhiji and every liberal out there.
Your candidate is ahead, finally.
There’s data and there are aberrations. To confirm a trend in the real world you have to look around. Check for common sense variations. Without that your “data” would be as good as our famously inaccurate opinion poll industry. You cannot just filter your data and pare it down to a certain specification until you find the result that you are looking for.
That’s called looking at aberrations, not data.
The good news for both liberals and opinion pollsters is that bad data has never affected their social, economic or political standing. So you can carry on.
UPDATE: After this article was published by OpIndia, we received a communication from Business Standard informing us that in wake of these details, they have withdrawn the story. The same can be confirmed by visiting their website here.