Transcript How Temperatures in the U.S. Affect Public Interest in Global

The Seasonality of Belief:
How Temperatures in the U.S. Affect Public
Interest in Global Climate Change
S. J. RALSTON
EAS 4480
Goals and Motivation

Wanted to investigate how the temperature
outside affected people's belief in global
warming


Since we can't directly measure belief, chose to
measure interest instead
Motivated primarily by comments made during
"Snowpocalpyse" this past winter

e.g. "Look at all this snow! So much for Global
Warming."
Methods


Obtained regional search data from the U.S.
from Google Trends
In the process of obtaining U.S. temperature
data from the National Climatic Data Center, a
department of NOAA


They don't want to give me the data and haven't
explained why yet
If no temperature data can be obtained, will
simply compare search data to season, or a
temperature proxy
Methods, cont.

Search data is provided on a weekly basis, but
temperature data (if/when it arrives) is on a
monthly basis



i.e. The search volume is measured once per week,
while the temperature is measured once per month
Therefore, interpolated the temperature data for
once-per-week sampling frequency
Examined search terms "global warming" and
"climate change" versus temperature using
Welch's Method
Methods, cont.


Looking at periodogram, phase lag, and
coherence of the data
Until real temperature data comes in, using a
sine function oscillating between +70 and +30
degrees Fahrenheit with a period of 6 months.


Values of +70/+30 were chosen based on national
averages from
https://www.ncdc.noaa.gov/sotc/national/
Search data covers a range of 120 months
(from Jan. 2004 - Dec. 2013), so using
matching temp. data
Search Volume: A Caveat

From Google Trends:


"Numbers represent search interest relative to the
highest point on the chart. If at most 10% of
searches for the given region and time frame were
for "pizza," we'd consider this 100. This doesn't
convey absolute search volume."
In other words, we know only how much interest
there was in "pizza" relative to the time there
was the most interest in pizza.
Google Trends: Relatively
From www.google.com/trends
Google Trends: For Reference...
From www.google.com/trends
Preliminary Results


Although temperature data has not yet been
acquired, an early look at periodicity—coupled
with a false temperature vector—has been
informative
Seems already to have a six-month periodicity

The question (which we will answer with the real
temperature data) is whether high or low
temperatures are correlating with the changes in
search volume, and if we can improve upon the
coherence observed
Preliminary Results, cont.
X: weeks, Y: temp and search volume
X: 1/weeks, Y: cross-PSD estimate
Preliminary Results, cont.
X: 1/weeks, Y: Coherence Estimate
X: 1/weeks, Y: PSD Estimate by FFT
Preliminary Results, cont.



Phase lag

For "global warming," up to +0.5 weeks

For "climate change," up to +0.5 weeks also
Significant periodicity observed at 6 months (27
weeks, to be exact), but this is likely due to
overbearing influence of regular sinusoidal
temperature proxy
FFT analysis of search terms alone yields a lot
of noise, vague periodicity around 6 month, 1
year marks
Conclusions

More robust analysis is needed



Specifically, real temperature data, if we want to get
anything useful
Preliminary results indicate a possible
correlation between seasonal temperature
changes and public interest in global climate
change
Pizza is way more interesting to the general
public than global climate change
THANK YOU!
Questions?