Under the Weather? How Social Media Emotions Reflect Weather Patterns
Grey skies getting you down? Research suggests that weather may impact our emotional state. In a new PLOS ONE study, Patrick Baylis from the University of British Columbia, Nick Obradovich from MIT, and colleagues wanted to find out if specific weather conditions are associated with the positive or negative feelings expressed via social media.
The researchers gathered 2.4 billion posts from Facebook and 1.1 billion from Twitter between 2009 and 2016. They used a categorization tool to analyze the sentiment for each post based on its positive and negative keywords. They also examined weather data for the location and date of each post to look for any associations.
Previous research has identified a potential link between weather and people’s emotional states, but which specific weather conditions trigger positive or negative emotions and how to measure these sentiments in an accurate and consistent way required further investigation.
[Panel A depicts the relationship between daily maximum temperatures and the rates of expressed sentiment of approximately 2.4 billion Facebook status updates from 2009–2012, aggregated to the city-level. It plots the predicted change in expressed sentiment associated with each maximum temperature bin. Panel B depicts the relationship between daily precipitation and the rates of sentiment expression of Facebook status updates. Panels C and D replicate these analyses for nearly 1.1 billion Twitter posts between 2013 and 2016 aggregated to the same cities. Shaded error bounds represent 95% confidence intervals calculated using robust standard errors clustered on both city-year-month and day.]
The scientists found that sentiments expressed on social media were strongly associated with certain weather conditions, whether positive or negative. Perhaps unsurprisingly, rainy, humid and cloudy days were associated with expression of negative sentiments. Temperature was also important: expression of positive sentiments increased with temperature up to 68 degrees, but declined once the temperature surpassed 86 degrees.
Days with a humidity level of 80% or higher were associated with negative expressions, as were days with a high amount of cloud cover.
[From Twitter, panel A depicts the relationship between daily maximum temperatures and the rates of expressed sentiment for non-weather posts, aggregated to the city-level. Panel B shows the relationship between daily precipitation and the rates of sentiment expression of non-weather posts.]
“We find that how we express ourselves is shaped by the weather outside,” says Nick Obradovich. “Adverse weather conditions — hot and cold temperatures, precipitation, added humidity, and increased cloud cover — reduce the sentiment of human expressions across billions of social media posts drawn from millions of US residents.”
Understanding the potential impact of weather on our emotions is important considering our constant exposure to weather conditions. The sentiment analysis tool used is imperfect, as its difficult to draw conclusions about how people actually feel from the sentiments expressed on social media. Nonetheless, this study may provide insight into how different weather conditions affect our emotions, which can act as a proxy for underlying human emotional states. Just another reason to hope for sunny skies!
[Comparisons between the effect size of below freezing temperatures on positive, non-weather, expressed sentiment with the effect sizes of other locale-specific events over the course of our data on the same sentiment metric at the Twitter city-level. The effect size of freezing temperatures compares in magnitude to other significant events. Error bars represent 95% confidence intervals calculated using heteroskedasticity-robust standard errors clustered on both city-year-month and day.]
Edited for WeatherNation by Meteorologist Mace Michaels