Biden vs. Biden. Google Trends

The election is in a month, give or take. America and the world are concerned about the outcome. Who will win the US election? The man in office, or Joe Biden? Polls are done by the dozen. Here is a “summary” of polls dated last Sunday 9/27:

Fig. 1: Biden-Trump voting intentions in National Polls. 9/27. Source: Financial Times.

The blue line in Fig. 1 plots voting intentions for Biden across time, 7 points above Trump (Red line). Careful, depending on the sample size, the margin of error can be around +/- 3.099, so a 7 point difference can be barely enough. As an example, if the “real” percentage for Trump is 43.1% + 3.099 = 46.2%, and “real” percentage for Biden is 50.3% – 3.099 = 47.2%, then the two candidates are somewhat close. A ten-point difference would be “safer”.

Bear with me. I won’t put formulas in the rest of this post. I will write as clear and simple as possible. The graphs are clear and easy to read, hang with me.

Now the complexity of US elections is beyond reason: Hillary Clinton won the popular vote, and lost on the electoral college. Let’s see what the Financial Times (a “somewhat” reliable source in my humble opinion) has to say. See fig. 2:

Fig. 2 Projected electoral college. 9/27, source: Financial Times

Apparently, things look good for Biden: 255 projected electors, 15 short of the required majority of 270. However those 255 include “leaning” states, and there still are ten “toss-up” states, including many conservative Southern states. If Republicans win all toss-up states, they gain a majority in the electoral college…

Many (on all sides) still doubt polls. Don’t. Having made an – honest – living in Market research for 30 years, I can personally vouch for pollsters. I know a few and trust them. However, there are two technical limitations to the accuracy of polls:

One, I already mentioned above, is the margin of error. When candidates are close to 50% each, the survey’s margin of error is highest. And the final result may be different from the poll figures.

Two: in many recent polls, a significant portion of voters for the “extremes”, e.g Le Pen in France or Trump in the US (okay, let’s plug in Bernie Sanders and Jean-Luc Mélenchon, so I’m not accused of being partial) tend to lie blatantly to the pollsters. They say: “I don’t know.” There can be as many as 15-20% of “I don’t know’s” in some polls. Enough to swing the balance on Election Day.

So, given the limitation of polls, let’s turn to another source: Google Trends. See fig. 3:

fig.3 Interest for Clinton and Trump, 2016. Source: Google trends.

Fig. 3 plots the “Interest” curves for Clinton and Trump during the 3 months before the 2016 election. While polls mostly gave Clinton the lead, “Interest” was always highest for Trump. Remember who won?

What is “Interest”? “Interest” is a fascinating concept! Google has been measuring for years, the volume of people who search/Google any word in any language in any country. For instance, how many people, day by day, have typed “Jojo rabbit” in Google, in the world or in the US? I suspect far more than “Marcel Proust”. Interest simply measures – and compares – the number of people who search information on Clinton or Trump. The number of people who type “Joe Biden” in Google. Simple and mind-blowing.

Interest is a very powerful tool. I have used it to measure the impact of marketing strategies, for brand names, brand image, etc. Cross-reference with advertising spending, on-line com, what have you, it works. Very high correlations. Now, in the case at hand, where do the two candidates in the 2020 election stand in terms of “Interest”? See fig. 4:

Fig. 4 Interest for Donald Trump vs. Joe Biden. 9/27. Source: Google trends

The volume of Google search for “Donald Trump” is 55%, way above the volume of search for Joe Biden, 43%. Many more people type “Donald Trump” in Google than they do “Joe Biden”. There is more “interest” for the former than for the latter. As in 2016, the red candidate generates more “interest” than the blue candidate. Anyone can access those results by typing “Google trends US elections” in Google. Quite different from the polls. What I don’t know, is how Google calculates 55 and 43%. Probably a sliding average. Let’s explore a bit more detail in fig. 5:

Fig. 5: Interest for Joe Biden and Donald trump, in the US, past 90 days.

Those curves have various points of interest. Trump (in red) led in interest in June-July, which probably explains his highest average of 55% Interest. Biden peaked on two occasions, one was the nomination of Kamala Harris on the ticket. Today, according to Google Trends, the two candidates are almost tied. Or seem tied. One key factor in using Google Trends (which I learned the hard way, working for corporate brands) is the word you type in. Interest results for “Coke” may be different from results for “Coca-Cola”. See fig. 6:

Fig. 6. Interest for “Joe Biden” and “Biden”, past 90 days.

More people search for “biden” (in red) than for “Joe Biden” (in blue). Logical. “Biden” is easier, quicker to type… Very little or no possible confusion with other names. I’ll spare you the “Trump” vs. “Donald Trump” graph. Same thing. “Trump” has a higher “Interest” (search volume) than Donald Trump. So let’s look at the Interest for “Biden” vs. “Trump”. Fig. 7:

Fig. 7. Interest for Biden vs. Trump, past 90 days, US.

Interest for Trump on September 26 (blue and red columns to the left) was 55, compared to 18 for Biden. If you go back to Fig. 3, Interest for Clinton vs. Trump, we have a similar situation as in 2016. Things don’t “look good” for the Democrats.

Conclusion: Google Trends does insist that “Search data is an indication of curiosity in the subject or candidate. It should not be considered an indication of voter intent.” (You can learn more by clicking on the link).

For example, one might Google “Trump”, despising the man. Of a Trump voter might Google Biden, looking for “dirt”. The New York Times spectacular release of Trump’s tax records might cause a surge in “interest” for Trump.

However the 2016 election precedent should cause concern in the Democrat camp. Hillary Clinton lost despite polls, because maybe her team didn’t know about Google Trends and/or did not take advantage of new, on-line forms of communication, while the other candidate tweeted his heart out every day. Obama won, amongst other things, thanks to innovative campaigning at the time: call centers, use of cell phones, texting, etc. The coming election in November can still be won by the Democrats investing heavily in on-line presence. Now! And, of course, if every Democrat goes to vote.

Thank you for reading this far, my blogger friends. I’ve been monitoring Google Trends for months now (biting my nails), wondering whether I should post this unusual text. I tried to make it as simple as possible. Don’t despair. There is still time.

You can do two things: one, go vote. That is a must. Two, share this. As much as you can. The key to winning this election is on-line presence and sharing. Plus, maybe someone knows someone who knows someone who knows… etc. Biden’s campaign manager. So they can take action. On-line action. There still is a full month.

My very best wishes for this coming election. (I look forward to the debate tonight)