Please Note: This project is not currently active. The content on this site is provided for reference and is not actively maintained.

Reading Twitter Users’ Sentiments On Gas Prices

by April 5, 2011

A couple of months ago, our team decided to dive into the discussion of gas prices on Twitter. We figured it would be a fertile topic and we weren’t disappointed. As the chart below illustrates, and as most Americans are well aware by now, gas prices saw a sharp increase in February 2011 after incremental increases over the past year. This in turn has prompted reactions on Twitter tweets ranging from resignation to ire to downright despair.

Similar to our exploration of weather tweets—see our discussion here—the topic of gas prices enjoys a kind of universality. Gas price fluctuations impact most of us weekly, even daily—and the reactions on Twitter bear this out. And as with our project on weather, emotional reactions to the topic have been the focus of this exploration.

We expected to see a rash of negative reactions to recent price trends. But how would the chatter on Twitter change in response to price fluctuations of different magnitudes? Different directions? How would the response differ regionally? These are the kinds of questions we hope to answer as we monitor reactions and overlay this information with existing gas price data.

But more than simply charting a dichotomous “high prices are bad” / “low prices are good” reaction, we hope to eventually explore more complex attitudes and opinions surrounding the topic. While negative reactions abound, many of these and other tweets provide revealing insights into a diverse range of linked topics: for instance, users’ assumptions on the factors that influence gas prices; behavioral changes prompted by fluctuating prices; and the use of foreign vs. domestic resources, among many others. See these tweets, for example:

 

But first things first. In our initial stages of topic development, we choose a topic, ask a simple question, ascertain that there is a sufficient volume of “chatter” about this topic in a given medium, and draft a short survey that we can apply to this medium to help us answer our question. (Kent lays out the major steps in our topic development process in this post.)

In the case of gas prices, our team had noticed that stories on gas prices seemed to be appearing more frequently in the news media, and using our Forecasting Tool we found that there was a fair amount of chatter about the topic on Twitter, too. This comprised our topic qualification phase.

For our topic validation phase, we drafted a list of keywords to help us yield a sufficient number of relevant tweets about our topic. After a bit of refining, we came up with the following list:

• gas gallon
• diesel price -jeans -sale -engine -coke
• unleaded gas
• unleaded price
• unleaded high
• unleaded low
• unleaded gallon
• #gas price -natural -propane -”nat gas”
• gas price -natural -propane -”nat gas”
• gas cost -natural -”nat gas”
• gas low -”low on gas” -propane -natural -”nat gas”

(Note: The words above that are prefaced by the minus sign [-] are words or phrases that were excluded from that keyword search because they tended to yield irrelevant results. Our forecasting tool automatically filters out any tweets that contain those words or phrases.)

Now, we already had a question we wanted to answer: namely, what emotions do Twitter users express about gas prices? At this stage, our team drafted several iterations of a survey incorporating this question, which we tested on a small sample of tweets. Initially, we judged whether the tweet was positive, negative, or neutral in respect to gas prices, but as the survey progressed we came to consider these neutral tweets as irrelevant—it’s the sentiment we’re really after.

The other important component of the survey is asking what kind of price the user is commenting on. Initially we asked whether the price in question was relatively low or relatively high. Eventually, we theorized that some of these reactions may be driven by the variability of the gas prices—particularly those that seem to change quickly—and so we added a third option of “variable” to encompass those tweets where the direction of the price is not immediately apparent, but its changing nature is.

These two components—emotion, or sentiment, and gas price direction—are the foundation of our survey, although it remains a work in progress. In the coming weeks, we will be testing the survey outside of our team for the first time, using distributed workers with our crowdsourcing partner, CrowdFlower. Look for more posts on these phases of our topic development!


Leave a Reply