What Are You Searching For?

Quantifying Political Differences with Internet Search Data

We all live in glass houses because we all use the internet. Others can see what we are searching for, and it isn’t always pretty.

It is often said that Democrats and Republicans have very different values and interests. Surveys are used to quantify such gaps, but they are often inaccurate and contradictory. Large sample sizes are cost-prohibitive, and people seldom reveal their true selves on surveys.

In contrast, internet search activity is nearly ubiquitous and does not lie. There are no fake searches. We look for things that interest us, and we do so in tremendous volumes. And our search activity is part of the public record thanks to services like Google Trends.

Research Method Overview

Google Trends shows popularity and trends for common search terms (aka “keywords”). It lets you compare states on search term usage. Scores from 0 to 100 are provided. If a keyword scores 100, then people in that state search for it more than in any other state. And the scores are relative so that you can compare high and low-population states.

I have been conducting political research with internet search data since 2017. After Trump’s shocking victory in 2016, I wanted to understand what drove the election and how pollsters and pundits could be so wrong.

I examined internet search activity for hundreds of keywords and reverse-engineered Trump’s win. I was able to answer this question: What did GOP leaders know about Americans that Democrats and pollsters did not?

Latent Culture War

The GOP suspected the US was ready for a culture war, so they launched one with shocking success. Trump and many other GOP leaders used cultural war slogans. Some of the most prominent Trump messages leading up to and during his presidency include:

• Misogyny: In the infamous 2005 Access Hollywood tape, Trump said: “They let you do it. You can do anything. Grab ’em by the p____.” (The tape recently helped the jury find Trump guilty of sexual abuse and defamation of E Jean Carroll.)
• Racism, Bigotry, Xenophobia: Trump announced his candidacy by accusing Mexican immigrants of being rapists.
• Racism and Bigotry: He criticized Black athletes (inferring to Colin Kaepernick) who knelt during national anthems.
• Anti-Democracy, Bullying: He promoted rough policing, telling law enforcement officers, “Please don’t be too nice.”
• Racism and Bigotry: He said he would never change military bases named for Confederate generals.
• Anti-Democracy, Bullying, Xenophobia: He said there were “very fine people, on both sides” rather than condemn the deadly violence in Charlottesville.

In hindsight, Trump successfully pushed our buttons of discontent. He forced it out into the open with his shamelessly uncivil behavior. It was a warped form of public service appreciated by many. He rallied the troops and won the presidency.

When examining search activity leading up to the 2016 presidential election, I found ample evidence for a latent culture war. Popular keywords in states that voted for Trump in 2016 were highly revealing, and I wrote several articles about it. I then used search activity to successfully forecast the 2020 Presidential and 2022 US Senate elections.

Solid-Red and Blue States

As we all know, our culture war rages on. But have the hot buttons changed? COVID is less relevant. Russia is waging war in Ukraine. Mass shootings are on the rise. Abortion rights are being eliminated. Voting rights are being restricted. Biden is older. Trump has been indicted. And so on. What messages will impact the 2024 election cycle?

In launching another forecasting effort, I start with 2020 and 2022. Specifically, what percent of the vote went to Democrats versus Republicans in these two cycles? I scaled the outcomes (via z-score), and a margin of safety was added for Democratic states to address electoral vs. population differences. Then the scores were combined, resulting in 15 solid-blue and 17 solid-red states. They are pictured on the map below.

It may surprise some that Texas and Florida are not in the solid-red category. What is often lost in the shuffle is that these large states have huge cities that are Democratic strongholds, such as Houston and Miami-Dade. Their impact cannot be gerrymandered away in state-wide elections. Hence such aggressive GOP efforts to restrict voting rights in these states. While the media tends to paint Florida and Texas as solid red, they miss the fact that many loyal Democrats live in these states.

Others may be surprised to see Minnesota or Michigan as not solid blue. But only 52.4% of Minnesota and 50.5% of Michigan voters chose Biden in 2020.

Keyword Selection

The next step is to find a wide set of keywords to test in these 32 states. Keywords are selected based on two criteria.

1. Keywords must have a minimum level of search activity in the US (average monthly organic search activity). At this time, the minimum is 10,000 searches per month.
2. A full spectrum of positive and negative keywords is pursued. For example, the Love category includes “love” and “hate.” The Abortion category includes pro-life and pro-choice keywords.

At this time, there are 754 keywords. This long list is possible because we are far from 2024’s election day. So I can use a 90-day timeframe in Google Trends. When it comes time to forecast the 2024 election, the timeframe will drop to weekly to detect any last-minute shifts in sentiment. Then, search volumes will need to exceed 50,000.

Keyword Filtering

The 754 political and cultural keywords are spread across 17 categories. The keywords are filtered for “polar strength” or how tightly they align with solid-red and solid-blue states.

To be polarizing, a keyword must satisfy one of two tests.

1. Most solid-red states have high interest in the keyword AND most solid-blue states have low interest.
2. The reverse of #1: Most solid-blue states have high interest in the keyword AND most solid-red states have low interest.

In the first case, the keyword is labeled “polar-red.” In the second case, the keyword is “polar-blue.” The first level of keyword filtering requires that at least 63% of the solid-red and blue states satisfy one of the two filters above. In other words, 20 out of 32 states must be polarized by the keyword.

The initial filtering output is summarized in the table below:

Overall, the 275 minimally polar keywords are nearly equally distributed between solid-blue and solid-red states (139 vs. 136). By category, though, there are significant differences. For example, the Lifestyle category has 11 more polar-red keywords than polar-blue. The opposite happens in Arts and Culture, where there are 23 more polar-blue keywords than polar-red.

Politically or Scientifically Correct?

As my upcoming articles will show, the 275 keywords provide tremendous insights into the dramatic differences between Democrats and Republicans.

Although many of the keywords are uplifting, many plumb the depths of hate, anger, and despair in our society. To shy away from these darker realities, though, is like sticking your head in the sand.

To say it another way, know your enemy. If Democrats are not aware of the underlying sentiments that drive people to vote Republican, then another 2016 is just around the corner.

Most importantly, when we understand what is driving our culture war, then we can take steps to heal and restore civil society. But you can’t fix what you don’t understand.

Next Steps

Over the following months, I will reveal the keywords within the 17 categories and discuss the stories they tell.