I’ve been seeing a lot of hoopla lately about people thinking search engines were biased for or against certain candidates. So, being a numbers guy, I decided to find out. I went to the 4 major search engines and tried a couple of searches. First, I searched for “2016 presidential candidates” on Duck Duck Go, Google, Yahoo, and Bing. Then, I searched for the candidate’s name plus the word “is”. The results are below.
Let’s set some ground rules.
- All of these searches were done on an iPad using Safari.
- All the searches were done from the site’s search bar (not the onmibox built into the browser).
- I was logged out of all search engines and I cleared my browser cache before starting this, so the results should be what a normal person will see on the public internet.
- All the screenshots in “2016 Presidential Candidates” search are “above the fold”, cropped to show only the top half of the screen, which is consistent with what a user would see on a tablet or phone with the keyboard visible.
- For the auto-complete suggestion comparison, I typed the candidate’s name plus the word “is” and looked at the auto-complete selections on each search engine.
- Rules for the stats: I put all of the auto-fill responses from each search engine into a spreadsheet and graded them on negativity. A “-” means negative, a “=” is neutral, and a “+” is positive. I then graphed the responses for each candidate and each search engine. To keep things fair, I graded things like insults as negative, I graded complements as positive, and anything else, like policy questions, age questions, or party affiliation questions as neutral. To get the “Favorability Score”, I gave each candidate a point for a positive result, and deducted a point for a negative result. Neutral results were not counted.
- Ignore the timestamps in the status bar. I’ve been trying a screenshot utility that randomizes time and battery status, and it’s not working very well.
DuckDuckGo
Duck Duck Go failed the first search for presidential candidates. 
“CandidateName is” Search on DuckDuckGo
Duck Duck Go was overall pretty negative. 
Overall Stats for DuckDuckGo
DuckDuckGo Stats
- Donald Trump wins (loses?) the negativity percentage on DuckDuckGo at 75%, and his negative results were the most extreme with references to hitler, a joke, and the antichrist.
- Hillary Clinton was second loser with 63% negative based on long-held beliefs that she’s a liar, evil, or a murderer.
- Jill Stein and Gary Johnson both had 50% negatives. Stein is apparently a hot, pro vaccine idiot loser. While “hot” may be a compliment and could be taken by some as a positive, I counted it as neutral because it’s not clear if it was offensive or not. Johnson’s criticisms are being an idiot joke moron. I counted the “being libertarian” questions as neutral policy questions, because no one knows if the users who skewed these to the top think that being a libertarian is a good or bad thing.
Google
Google surprisingly also failed the 2016 candidates search. 
“CandidateName is” Search on Google
Google was far more positive overall than the other search engines. 
Overall Stats for Google
Google Stats
- Trump got an “awesome”, but he also got “shape shifting lizard”, so we’ll call it a wash.
- Clinton was compared to both a man and Cersi Lannister, so I don’t know what to make of that.
- Stein has only one negative, “unqualified”
- Johnson has an “awesome”, but so did Trump, there’s that.
Yahoo
Yahoo did slightly better on the candidate search, but left out third party candidates and still lists Bernie Sanders, who bowed out weeks ago. 
“CandidateName is” Search on Yahoo
Yahoo returns pretty middling search auto-complete, as far as negativity vs. positivity. 
Overall Stats for Yahoo
Yahoo Stats
- Trump and Clinton both get 66% negativity ratings. Trump is an antichrist crazy idiot joke, while Clinton is an evil criminal liar devil.
- Stein and Johnson are idiots who are wrong. And Yahoo also thinks Stein is hot. Weird.
Bing
Bing easily wins the 2016 Candidates search. It returns the four people running nationally, including their VP picks. 
“CandidateName is” Search on Bing
Bing’s results had fewer overall negatives, but the negatives were a lot more vicious. 
Overall Stats for Bing
Bing Stats
- Trump had only two negatives, but they were “malignant narcissist” and “lying racist”
- Clinton also had 2 negatives, both calling her a liar
- Johnson and Stein had similar results. Mostly policy searches, but a little bent towards fraud or dishonesty.
Conclusions
What does this all mean? We don’t really know. I can’t see any pattern of a specific search engine skewing results towards or against one candidate or another, as is being suggested all over Twitter. 
One big conspiracy I’m hearing is if you Google “Who will win 2016 election” you see Hillary Clinton. I tried this in all 4 search engines and did not see the same results. Again, these are above-the-fold screenshots. Google did come the closest to the above accusation after scrolling down, but it was in the form of a table that showed latest odds from predictit.org.
I just don’t see any search engine skewing search results, even a little.
If you are seeing it, tweet me some screenshots, and hashtag it #SearchBias. I’d love to see it.
|||