Earlier this month, we released an interactive piece of content exploring search and online media around the US election. With the United States of Search we looked at what people are searching for in Google, and also the websites and pages they are finding as a result. So what does this search data reveal about the US Election 2016?
Contents
How Much is Search Impacting the US Election?
Search is having massive impact on the US Election. From February of this year, search volumes for election-based searches increased to 69.3 million — in some cases, with volumes for individual keywords tripling on November-January averages1. It’s powerful evidence of the role that search is playing when it comes to the way voters are seeking additional information during a major political election.
Perhaps unsurprisingly, the majority of these searches are from people looking for information about the candidates themselves. Google Trends shows a significant increase in search volumes for ‘Donald Trump’ and ‘Bernie Sanders’ for example. As the following graph illustrates, all the Republican and Democrat candidates still running a campaign for office have experienced a notable increase in search volumes in the past few months.
This is a important development, for in politics, where battle lines are often drawn around such topics. Searches such as ‘Donald Trump immigration’, ‘Ted Cruz gun control’, or ‘Bernie Sanders marijuana’, for example, (and the content people subsequently find) likely act as influential touch-points in voter decision-making journeys.
And increasingly, the impact that search having on politics is being explored in greater depth. Earlier this month Justin Wolfers of The New York Times wrote on the strong correlation between surges of election day searches for a particular candidate and the share of vote that candidates were winning.
“It’s not a crazy idea. After all, who among us hasn’t searched for reviews of a car, a stereo or a phone on the day of buying it? And if we do this when we’re shopping, who’s to say that people don’t do the same on Election Day?”
- Justin Wolfers, The New York Times
It’s not an unfeasible proposition. Research increasingly shows that organic search is the channel of choice for consumers when researching or evaluating a purchase decision2 — and naturally there’s no reason why search shouldn’t be the channel voters turn also when they are looking to research or evaluate a voting decision.
Thus with the United States of Search, we sought to explore search and online media around the US Election beyond the typical studies of search volume and topical interest. Using the Linkdex platform we also looked to gain an understanding also of the content that voters are seeing, and one way or the other, are influenced by.
What is the United States of Search?
The United States of Search looks at the popularity of individual candidates, topics, websites, and pages, presenting a view of how search behaviour and content consumption varies state by state, and over time (each month we will update the resource with the latest search data, creating a historical record of search all the way up until the Presidential election in November).
We designed the resource to be an interactive tool allowing journalists and marketers alike to explore the data, find statistics, and discover their own stories. Across the resources’ many pages, you can view the data at a national level, segmented by individual states, or by any individual candidate (where data incorporates searches where a particular candidate is referenced).
In the meantime here are some initial insights we’ve gleaned from the March data.
March Analysis: Search Popularity of Parties and Candidates
Republicans vs Democrats
First things first — what does the United States of Search look like when we total search volumes for any search which references either the Republican or Democrat party, or any of the candidates that have run for either of the main parties.
The above visualization shows the percentage difference for any Republican/Democrat search from the national US mean. In other words, in Texas where 78.7% of searches are for the Republican party or any of their party candidates, we say the state has 4.1% percentage difference from the US average in favour of the Republicans (where the US average for Republican searches is 74.6%).
Essentially, this visualisation allows us to view the extent to which state specific searchers are more likely to search for something that references either the Republican or Democrat party, or any of their candidates.
(N.B. It’s fair to assume that Republican searches vastly outnumber the equivalent Democrat searches at the moment due to the competitiveness and heightened media coverage around the current Republican nomination schedule.)
Observations
- Clear increased search interest for Republicans in the so-called ‘Deep South’ (namely South Carolina, Mississippi, Florida, Alabama, Georgia, Louisiana, and Texas).
- Arkansas, Tennessee, and Kentucky also have greater search interest for Republicans than average.
- Much of New England has greater search interest for Democrats, especially so in the state of Vermont (Bernie Sander’s home state) which registers an astonishing 29.9% greater than average search interest, in the main due to keyword volume for the ‘Bernie Sanders’.
- Greater Democratic search interest in New England is with the notable exceptions of New Jersey, Pennsylvania and Maryland (typically considered to be strong Democrat states) which, at least in regards to search volumes, are leaning Republican.
- The Midwest and West generally has a greater search interest for the Democrats, with the notable exceptions of Idaho, Utah, and Arizona.
Analysis
Whilst it’s clear that search volumes do not correlate entirely with typical state voting preferences, to a large extent they do. Search in the Deep South, the Midwest, and the Western Pacific states are aligned with typical voting expectations3.
That search interest should be generally aligned to typical voting expectations is not unsurprising, and coincides with Google’s implication; that increased search interest for a particular candidate correlates directly with voting preference.
There are however, notable exceptions in New Jersey, Pennsylvania and Delaware — typically considered strong Democrat states — which especially in the case of New Jersey indicate increased search interest in Republicans.
Oklahoma, Wyoming, and Montana — considered strong Republican territory, also defy typical expectations, exhibiting a greater preference for Democrat related searches.
Are there any feasible explanations for these disparities? To get to the bottom of the matter, we took a closer look at what exactly people were searching for in these states.
Further analysis
On the first point, New Jersey exhibits a 3.9 percent greater than average interest in Republican search terms, and this can to a large extent this can be attributed to Chris Christie — the New Jersey Governor (who albeit recently withdrew from the presidential race). Christie draws a 4.5 greater search interest in his home state over national average, and this volume accounts to an extent for the state’s Republican leanings in search.
To a lesser extent, this effect can also be seen in neighbouring state Delaware, where searchers for the New Jersey Governor are 1.9 percent higher than the national average.
In Wyoming, Montana, and Oklahoma, which lean blue in search, we found that a great deal of the democratic search volumes were for the keywords referencing ‘bernie sanders’ and ‘hillary clinton’. Searches referencing Sanders were higher than average in Montana and Oklahoma, markedly so in the latter, where searchers for the candidate were 2 percent greater than the national average. Wyoming and Montana on the other hand exhibited greater interest in Hillary Clinton, by 2.3 and 6.6 percent difference over average respectively.
Swing States
Overall, with the notable exception of Wyoming and Montana, there certainly does seem to be a notable correlation between overall search interest and the expected voting behaviour of individual states (based on historical results).
Perhaps where the data can be most indicative is in swing states, where any indication of voter sentiment can be a valuable signal. The following chart illustrates their political leanings with regards to search:
Whereas Florida and Ohio are both indicating significant preference over average for Republicans, this is perhaps accentuated by the home state advantages of Marco Rubio (Florida) and John Kasich (Ohio). With neither candidate standing a realistic chance of winning the nomination, it is difficult to place any meaningful prediction based on search behaviour.
Iowa, with a less than 1 percent Republican preference, is perhaps too close to call, but interesting New Hampshire and Nevada both exhibit a slight Democratic leaning.
Where there is clear party preference within swing states is in Virginia, which exhibits a 3.9 percent preference for Republican searches, and Colorado which has a 7.2 percent proclivity for Democrats in search over the national average, the second greatest positive difference for the Democrats in any individual state after Oregon.
So could it be that search is indication that Virginia will vote Republican and Colorado Democrat come November? Whether or not search can predict overall voting behaviour remains to be seen in the coming months.
Predictions
- Virginia to vote Republican and Colorado to vote Democrat in the Presidential Election
Remaining Unknowns
Why are Montana and Wyoming exhibiting such high democratic search volumes? Wyoming is often cited to be the most Republican state in the US 4 so could this be a case of voter hapathy manifesting itself in lower search interest? Do states with heavy political leanings therefore underepresent voter sentiment in search?
The Primaries and Caucuses
Much of the political activity and coverage over the last few months has been centred around the Democratic and Republican primaries and caucuses. For the results that have already come in, we have the advantage of looking at the results retrospectively and seeing where, if at all, that results have correlated with search data.
The following chart shows search interest for Bernie Sanders, as compared to their overall share of vote in the elapsed state primaries to date:
The results show that to a certain extent a greater share of search volume does coincide with a greater share of vote within a state primary.
Home state advantages yield the greatest deviations to this measurement, and this reflects actual voting. Bernie Sanders won the Vermont primary with a 86.1% percent share of the vote, the greatest winning margin for any primary to date in the current electoral race. According to our data, people in Vermont are 23.5 percent more likely to make a search referencing Bernie Sanders than average, which was also the greatest margin when it came to search deviation for any state against the US mean.
The home state advantage can also be seen for John Kasich in Ohio, where the candidate was 4.7 percent more likely to be searched for than the US average, and where he won the state primary with 46.8% of the vote.
Kasich has been perennially trailing more than a few of his adversaries, and Ohio has been the candidates strongest state to date. However the Ohio Governor did managed a surprise second place in New Hampshire primary, winning 15.8 percent of the vote. This was reflected in search where search interest of 2.7 percent over mean represented the greatest share of search in any state except Ohio.
To a similar extent, Ted Cruz experienced similar heightened search interest in his home state of Texas, where people were 5 percent more likely to make a search referencing Ted Cruz than the national average. Cruz took 43.8 percent of the vote in the Texas primary to win the majority of delegates, so it appears that this interest translated into votes.
Interestingly, search interest for Cruz in Iowa was also well above mean, at 3.9 percent greater than the national average and this may have been fuelled by his victory in the states Republican primary in early February.
Clinton vs Sanders
When it comes to the battle between Clinton and Sanders, there are a few states in which one of the candidates is receiving a greater search interest over mean.
In Oregon, search interest in Sanders is 6.9 percent over mean, the second highest for any state other than Vermont. Could this be indicative of a strong Sanders turnout in the state primary on May 17th?
As touched upon earlier Clinton receives a surprising amount of search interest over mean in Wyoming (6.6%), Montana (2.3%), and North (3.9%) and South Dakota (3.3%). Wyoming is the only of these states to have held a primary/caucus to dater, and it actually went to Bernie Sanders — who performed less well than Clinton in search — where people in Wyoming were 3.3 percent less likely to make a search referencing Bernie Sanders than US mean.
In this instance share of search interest does not preclude voting behaviour. However, if forced to go out on a limb one could say that if Clinton could not translate strong search interest in Wyoming into votes, that she is even less likely to do so in neighbouring midwestern states — meaning, that they could well go the way of Sanders.
New York and California
Of the remaining primaries, New York and California are highly sought after by all the remaining candidates from either parties. For both the Republicans and the Democrats, the battle will a direct battle between two candidates.
Trump vs Cruz
If results of the previous two months are anything to go by, then Trump will be victorious in both New York and California — Cruz is perhaps too conservative to win popular support in states with predominantly urban populations. If the voting does go this way, results will correlate with search interest in these two states. Trump receives 2.6 percent in search interest in California, and 1.1 percent in New York over US mean. Cruz on the other hand, receives 1.1 percent less in California, and 1.4 percent less in New York.
In the Democratic primaries, the battle is harder to call, for although Sanders seems to generate a greater proportion of search interest in these two states, the states he has won to date have been those in which he has generated significantly higher search interest. Sanders does have 2.7 percent greater search interest in California, and 1.6 percent in New York. Clinton has ‑0.1 percent search interest in California, and 1.2 percent greater than average search interest in New York.
If we take Wisconsin as a precedent — where Sanders had 1.2 percent greater search interest, and Clinton ‑0.7 percent. Sanders took a narrow victory in this state — suggesting he could do the same in California. With his advantage less in New York, we’re going to have to give the state to Clinton.
Predictions
- Trump to take New York and California convincingly.
- Sanders to win Oregon, Montana, and North and South Dakota.
- Sanders to edge a close battle over Clinton in California.
- Clinton to win New York.
Final Thoughts
Admittedly this is far from a scientific study of search and politics. The above analysis is intended to be a perspective of current progress in the US Presidential Election — as informed by search. Whilst it has unearthed some of the intricacies between search interest and voter preference but while there are some correlations, it clear there is also great deal of unique nuance here.
Overall, search interest can be viewed as an indicator of voter sentiment — to an extent. It’s worth noting that whilst our data is in many respects, reasonably comprehensive, certainly in its keyword and geographic scope, it doesn’t take account of potentially key data-points such as date or time of day.
Stay tuned for further analysis on the United States of Search. Next time we’ll be exploring topics, websites and page visibility.
Research and Methodology
Our research began by collating a keyword list to cover a full spectrum of political searches in the US. This included all candidates from either the Republican or Democrat party to announce a campaign in November 2015, political search terms such as ‘us election’ or ‘swing states’, and a full range of political topics and issues ranging from ‘gun control’ to ‘health care’, ‘climate change’, or ‘islamic state’, to name a few — and also included combinations of these terms such as ‘ted cruz gun control’.
Search volumes were then obtained from publicly available Google search data for the period between November 2015 and January 2016 segmented by each individual state. Keywords were then tagged and categorised into relevant entities, so we could see the volume of searches being conducted for any searches referencing either the Republican or Democrat party, any individual candidate, or any individual topic or issue.
To find out what content searchers were visiting as a result of their searches, we used the Linkdex SEO platform to ascertain which websites were winning share of search for any of these keyword groups, for each individual state. Following, are a few examples of how the data can be compared and cross-referenced (and I think you’ll agree with some interested results).
You can find out more about the methodology behind our research here.
- According to our defined keyword lists and based on publicly available Google data. ↩
- http://www.edelman.com/insights/intellectual-property/2016-edelman-trust-barometer/global-results/ ↩
- http://www.gallup.com/poll/188969/red-states-outnumber-blue-first-time-gallup-tracking.aspx ↩
- http://www.gallup.com/video/181496/trend-line-wyoming-utah-republican-states.aspx ↩
Contributors
Pat Hong
Editor at Linkdex/Inked, Linkdex