Seth Stephens-Davidowitz, "Fifty States of Anxiety," New York Times, AUG. 6, 2016
image from article [Above U.S. Average: red +10% or more; orange +5 to +10; pink: 0 to + 5'; below U.S. average: light violet: 0 to -5; medium violet: -5 to -10; violet: - 10% or more]
Feeling worried? These days, much of America is.
Over the past eight years, Google search rates for anxiety have more than
doubled. They are higher this year than they have been in any year since Google
searches were first tracked in 2004.
So far, 2016 has been tops for searches for driving anxiety, travel anxiety,
separation anxiety, anxiety at work, anxiety at school and anxiety at home.
Americans have also become increasingly terrified of the morning. Searches for
“anxiety in the morning” have risen threefold over the past decade. But this is
nothing compared with the fear of night. Searches for “anxiety at night” have risen
For years, I have confidently pontificated on topics that I think are important
but that I have little experience of — child abuse, racism, sexism, sex. Now I am
ready to tackle a topic I actually know something about.
Over the past few weeks, I’ve taken a break from worrying about my own
anxiety to studying our country’s. While I am not sure I totally nailed down why
anxiety seems to have risen so much during the Obama era, I did learn a lot.
The places where anxiety is highest are not where I would have expected. When
I was growing up, if you had asked me which people were the most anxious, I would
have said New York Jews. And a decade of interacting with our country’s urban
intelligentsia, Jewish and otherwise, has confirmed that pretty much all of us are a
But data has now given me a window into how the rest of the country lives.
They’re even worse off. Google searches for anxiety tend to be higher in places with
lower levels of education, lower median incomes and a larger part of the population
living in rural areas. While the state of New York does have above-average anxiety, it
is actually higher in upstate New York than it is in New York City.
Searches for panic attacks are overwhelmingly concentrated in less educated, poorer
parts of the country, particularly Appalachia and the South. Test-taking anxiety is
highest in Arkansas. Searches for “anxiety about death” are highest in Kentucky.
The epicenter of anxiety, according to Google, is Presque Isle, Me., where fewer
than 20 percent of adults hold a bachelor’s degree.
An important note about the data: Google’s measure of anxiety includes a broad
range of searches, including “anxiety help,” “anxious” and “anxiety symptoms.”
While Google searches may not be a perfect measure of anxiety, there is increasing
evidence that searches on a health condition highly correlate with the number of
people suffering from that condition. The rates of Google searches for anxiety in a
state also correlate with survey measures of anxiety.
The data also suggest that some things I thought for sure would be huge drivers
of anxiety turn out not to play much of a role at all. One example: terrorism.
I tested how anxiety searches changed after the five largest recent terrorist
attacks in the European Union — Madrid, London, Paris, Brussels and Nice — each
of which killed or injured more than 100 people. In the weeks that followed, there
was no rise in anxiety searches in the country that was attacked. I also tested
whether anxiety searches rose in Boston after the marathon bombing. They didn’t.
Ditto for Orlando, Fla. after the nightclub shooting.
Of course, it may be that terrorist attacks terrify people without making them
Google about their anxiety, but it still surprised me.
Another potential source of anxiety that doesn’t pop up in the data: Donald J.
Trump. Many people, I assumed, are anxious over the prospect that Mr. Trump
could become president of the United States. And I thought they would express this
anxiety through Google. However, while anxiety searches have continued their
decade-long rise this year, there was no noticeable change in this trend as the
prospect of Mr. Trump’s actually winning became more plausible. Nor have
Democratic-leaning parts of the country witnessed an unusually large rise in anxiety
I also thought Mr. Trump’s fear-mongering was helping to create fear. Anxiety
must have risen after Mr. Trump’s dystopian acceptance speech at the Republican
convention last month, right? Not so. There was no increase in searches related to
anxiety or panic attacks.
In fact, the data suggest that fear-mongering politicians may not create all that
much anxiety. Consider the 2004 presidential election between George W. Bush and
John Kerry, in which the incumbent continually emphasized the threat of terrorism.
I compared anxiety searches in the summer and fall of 2004 in swing states and
non-swing states. Swing states, where advertising was concentrated, would have
been exposed to far more fear-based messages. But they did not show more anxiety
or panic attacks.
I did find two factors that cause anxiety. The first — and this is hardly a surprise
— is a major recession. States that were more deeply affected by the Great Recession
saw bigger increases in anxiety during and after the recession. I estimate that each
percentage point increase in unemployment is associated with a 1.4 percent increase
The second was not something I had thought about, so let me explain how I
even stumbled upon the possibility. Google Correlate is a service that makes data
available to researchers so that they can test which searches are most frequently
made in the weeks a particular search is made. For example, at certain times of year,
when people are searching a lot for “sweaters,” Google Correlate tells us they are also
searching for “scarves,” “fleece pajama pants” and “hot cocoa.”
A few weeks ago, I put “panic attack” in Google Correlate, and one of the highest
correlated search queries was “opiate withdrawal.” Panic attacks are a known
symptom of opiate withdrawal, although looking at the striking correlations here, I
wonder if they may be an even more common symptom than we had realized.
The places with high opiate prescription rates — and high search rates for opiate
withdrawal — are among the places with the highest search rates for panic attacks.
These areas include Appalachia and the South.
During years in which many people complained of opiate withdrawal, many
people also complained about panic attacks. Interestingly, searches for opiate
withdrawal and panic attack have continued to rise over the past few years, even as
opiate prescription rates have finally fallen.
Searches for opiate withdrawal consistently start high at the beginning of the
year. They mostly drop through the year, although they rise in the summer and then
surge around Christmas. New Year’s resolutions and doctor vacation schedules may
play roles in these patterns. Searches for panic attacks follow a similar pattern,
although there is less of a rise around Christmas.
To be honest, I still don’t know what to make of this huge rise in anxiety-related
searches. I have a lot of questions. Has fear of a black president contributed? Is
anxiety contagious — if someone loses her job and gets anxious, are her kids and
friends likely to become more anxious as well? Is anxiety subject to tipping points
and bubbles? How much of a role is the return of heroin playing in the rise of panic
Before I look into this further, though, I will take a break from working and
instead indulge in my own favorite 2016 habit. I’m going to pace the streets of New
York and fret.
Seth Stephens-Davidowitz is an economist and a contributing opinion