Quantitative journalism with open data

This is the best news article I’ve seen in a while:

It’s the political cure-all for high gas prices: Drill here, drill now. But more U.S. drilling has not changed how deeply the gas pump drills into your wallet, math and history show.
A statistical analysis of 36 years of monthly, inflation-adjusted gasoline prices and U.S. domestic oil production by The Associated Press shows no statistical correlation between how much oil comes out of U.S. wells and the price at the pump.

Emphasis added. It’s a great example of quantitative journalism. They took the simple and oft-repeated statement that increased US oil production reduces domestic gas prices (known colloquially as “drill baby drill”), and they subjected it to a few simple statistical tests for correlation and causality. The result is that there is no correlation, or at least not one that is statistically significant. They tested for causality using the notion of Granger causality, and they found that if anything, higher prices Granger-causes more drilling, not the other way around!
And here’s the very best part of this article. They published the data and the analysis so that you can check the numbers yourself or reach your own conclusion. From the data, here is a scatter plot between relative change in price per gallon (inflation adjusted) and the relative change in production:

What’s more, they asked several independent experts, namely three statistics professors and a statistician at an energy consulting firm, and they all backed and corroborated the analysis.
Kudos to Jack Gillum and Seth Borenstein of the Associated Press for this wonderful article. I hope we can see more examples of quantitative journalism like this in the future, especially with open data.

8 Replies to “Quantitative journalism with open data”

  1. This is welcome, but ‘quantitative journalism’ that doesn’t investigate underlying logical premises is not much better than lying with stats.
    The fundamental issue with drilling and the environment is this – any unit of oil consumed, independent of how much is consumed in total, or whether this amount is going up or down, has to come from *somewhere*. If it’s not coming from the Atlantic coast it’s going to come from the North Sea, or Canada, or Venezuela or Saudi Arabia. Wherever it comes from, drilling causes ecological trouble, but given (and a fortiori since the US uses a quarter of the world’s oil) that the oil is consumed, there is no *environmental* argument against drilling in the US, since the US isn’t unique or even exceptional in having ecosystems. The whole environmental (unfortunately “liberal”) argument against US drilling for oil is therefore a huge red-herring, and amounts to nothing more than green varnish given to NIMBY-ism.
    Given that AP isn’t really scrutinizing the ethical and logical basis of the green anti-drilling argument the way it does the pro-profit motives of oil companies, just adding numerical razzle-dazzle isn’t all that useful. Indeed, since you get to beat people over the head with (Science!!) numbers they don’t understand, the strategy effectively just uses numbers as ammunition in a political battle.

    1. I agree that it is important to dig deeper into the underlying logical and ethical premises for the different positions on drilling. But that isn’t the point of my post, nor the article. If there is one side of a debate (either side!) that holds as fact something that is easily countered through a simple statistical analysis, then this should be published in a clear and unequivocal rebuke. Adding a deeper discussion of the pro/con for drilling would actually weaken the article because it would remove focus from this one powerful point.
      While you bring up a good point about the dangers of lying with statistics, I think these reports did a good job by asking several independent experts to verify the study, including an energy consultant. This adds a lot of credibility. So there are ways to counter this lying effect.
      And suppose that there were a significant correlation, or even causation. Fine. This just means that among the arguments against drilling, price insensitivity isn’t one of them.

      1. Steve, thanks for the response. I agree that the analysis seems mathematically sound, and it’s not my intention to say fact-checking is bad. I’m not saying the AP article is wrong. I was pointing out what AP did not say, and indeed has never said to my knowledge. It would be far from difficult for AP to do a series comparing ecological damage between drilling off Florida and drilling in Venezuela. There would be plenty of facts there too, and they’d dismantle the US green view that drilling here specifically hurts the ecosystem.
        Fact checking too can be an instrument of politics if deployed selectively, as I believe it is being here. I underline, my point is not that facts are bad, more like, if AP picks facts and stories that support one side while omitting those supporting the other, it’s hard for me to regard this as a contribution to the public discourse, independent of how impeccable the math in the story is.

    1. This is the second time you’re missing the point of this posting. All the original AP article shows is: “The argument in favor of (more) drilling that it will reduce the gas price for consumers is not substantiated by facts.” It does not say anything about other arguments.

      1. They were two consecutive posts, made not two minutes apart, actually 🙂
        My point is *not* that fact checking statistics is bad, but that such fact checking itself can function as a political stick if you only fact check where it supports your political preferences. As here, I claim.

  2. There was a related “EconTalk” episode on this a few weeks ago. They talked about various policies related to energy use and how the effects are often not what people naively assume they would be:
    It seems to me these sorts of examples demonstrate the usefulness of the kind of analysis performed by economists. You don’t just look at how changing one variable effects the system with everything else held constant. You look at whole the whole system — in particular, the equilibrium of the system — shifts when one thing is changed. The equilibrium may shift in the opposite direction of what you would think due to how all the other variables change in response.

  3. In response to Steve’s last point, I don’t know about seeing more quantitative journalism, but you can definitely listen to it on the excellent BBC radio program (or should I say programme?) More or Less, which according to its webpage “is devoted to the powerful, sometimes beautiful, often abused but ever ubiquitous world of numbers.”

Leave a Reply

Your email address will not be published. Required fields are marked *