« Trifacta revisited: tackling a Big Data problem | Main | How Optimizely will kill your winning percentage, and why that is a great thing for you (Part 1) »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Tom West

It depdns what happens to that $100 after you spend it. With high gas prices, it goes into oil-producing nations' blank account, and generally sits there. With low gas prices, it probably ends up in some buisness that then can employ more workers, who can go and spend more money, etc. So that $100 will get spent multiple times, and it's *that* which counts as increased economic activity.


Which is why GDP is a completely silly measure of anything (as are most Keynesian 'measures').

"The GDP framework gives the impression that it is not the activities of individuals that produce goods and services, but something else outside these activities called the "economy." However, at no stage does the so-called "economy" have a life of its own independent of individuals. The so-called economy is a metaphor—it doesn't exist."



Tom: Thanks for the further explanation of the economic reasoning. It seems like there are other required assumptions such as that everyone will in fact spend the $100, and that where the money goes, it doesn't still lend in somebody/nation's bank account.

Nate: I agree that the government's use of GDP is problematic - was going to write about it in the Numbersense book but decided to talk about CPI and unemployment numbers instead. Also, in the first two chapers of Numbersense, I argue that most metrics "don't exist", and are "subjective"... and I mean everything from ranking schools to Google pagerank to defining a company's "profit" so for me, to say a metric is bad because "it is a metaphor" is to say almost every metric is bad. The "better" metric will still be a metaphor. The bigger problem is not the metric, but the inevitable subversion of it. This is all in Numbersense.


I'll check out the book. I'm not against metrics completely. But it's not just the subversion of them you need to watch out for - it's the replacement of *real things* with metrics. Take a simple metric in the fast food industry - "time from order placement to delivery". This metric, by itself, tells us nothing about food quality, order correctness, employee morale, etc, etc. Relying on it to make decisions might lead to a long run disaster. At least this metric only claims to measure one thing.

The aggregate economic metrics are much worse, because they make incredibly broad claims about what they measure and how that measure is useful. Not only that, but they don't include any measurement error, and aggregate things that in many ways can't be aggregated together. How can one compare an LED TV with a Refigerator? Price is a function of time, geography, and the people involved in the transaction. Don't even get me started on the chain weighting mess that is GDP "measurement".

The fundamental truth is that it is impossible for us to measure something as complex as the trillions of interactions between people exchanging goods for goods, services for services, or a bunch of less-tangible exchanges involving currency, goods, or time. When we try, we are inevitably led into the belief that with "proper policy", we can *direct* the outcome. Which is, of course, the fatal conceit.


Nate: great points. I take a less fatalistic approach. There are times when you need aggregate metrics. A policy maker has blunt instruments; there is one law for everyone so some kind of aggregation is necessary. The key is to understand the assumptions behind the averaging, and make sure they are appropriate for the specific application.
A nice counterexample is google pagerank. The so-called authority of a web page is an invention and is affected by lots of factors. For the purpose of a search engine, pagerank is an accepted aggregate metric. If one wants to nitpick, there are many things that pagerank fails to capture.
Reification is a huge problem. That's one reason for my recent post on "proxy unmasking". In online banner advertising there is a metric called "viewthrough". Except that it does not measure viewing ads, it measures serving of the ad to a browser. Not the same!


You miss the effect of constraints and potential multipliers when looking at substitution. Let's say households are constrained by their resources and would direct every dollar saved from lower gas prices to other needs. The question for the basic GDP calculation becomes one of multipliers: does buying gas have the same multiplier as buying food or clothing? I don't know the calculations but there are different levels and types of multipliers based on industry.


jonathan: that's what Tom also said in an earlier comment. I find it to be the case with much of economic reasoning - that arguments are based on second-, third- or fourth-order effects. Often the first-order effect appears to be small or even reversed. It would seem difficult to find data to support higher-order effects. I'm not an expert in this area; how is it possible to know the values of multipliers by industry at different levels? I mean, without making lots of assumptions.


I generally like your posts, but I was disappointed by this one.

* The issue of multipliers might be a second-order effect, but it doesn't mean it's a subdominant effect when considering the impact on the overall GDP. This isn't a perturbative expansion -- the higher order effects matter.

* The US is a net importer of oil. So most of the $100 that would have been spent on oil, in your example, would have effectively left the US, with no multiplier within the US economy. This isn't a case of McDonalds and Burger King across the street from each other, it's more like buying a domestic car instead of an import. (Though even that example is not straightforward, depending on where the import was made.)

It's fine to be skeptical, but you're claiming these stories "are likely wrong", which goes beyond skepticism. They may be misleading in suggesting a more direct link than actually exists, but that doesn't make them wrong, just too simple.

I agree that the journalists' job is to establish the link between oil prices and GDP in a way that both captures all the causal relationships and is clear to the reader. But if we called every story written by a poor journalist "wrong" then we'd often be throwing the baby out with the bathwater.


In the Burger King vs. McDonald example, I agree there is no change in GDP assuming the velocity of money stays the same between the two which is probably a safe assumption (GDP=money supply x velocity). The difference with gas is that a portion of that is the price of crude. Based on the EIA website, 62% of the price of a gallon of gas can be attributed to crude prices (as of November 2014). So if we imported 100% then 62 cents per dollar would be taken out of the money supply. However, based on the EIA data, net petroleum imports as of Q1 of 2014 was 28.5%. So really only about 18 cents per dollar spent on gasoline leaves the country (assuming none of it comes back to the US). Over the past year, oil prices have dropped by 50% so the value or contribution to GDP of the 71.5% produced in the US has also been cut in half. So from the GDP point of view, the question is which is larger, the 18 cents per dollar taken out of circulation or the 50% drop in oil prices from US production. According to EIA, US consumed 19 million barrels/day of oil. Of that 71.5% or 13.6 million barrels/day was produced in the US and 5.4 million barrels/day was imported. If the price of oil drops by 50% then the loss of US GDP from oil is equal to the value of 6.8 million barrels/day. And that is greater than the 5.4 that is imported (money leaving the country). So based on this back of the envelope calculation it seems to me that US GDP will decrease and not increase as all the talking heads have been reporting. Any thoughts, refinements on my back of the envelope calculation?


Mike and PC: welcome to the conversation. I think PC's response is exactly the kind of math I am looking for. I think it is the journalist's job to sort out the facts, not to repeat some chain of logic that may or may not work out.

Mike: the headline of my post say "three reasons to doubt..", not "three reasons why this is wrong". The point of the post is that I want to see some actual data to support those arguments. Even in PC's short calculation, there is a whole series of assumptions along with the data, for which any responsible economist/journalist should provide. I am not an economist but I'm not willing to just take this difficult question and accept a simple answer.

PC: My point #3 is that this type of calculation now assumes the current price level is the new normal. I think it is likely to be a local minimum, and chances are the price will not stay at this level for the entire year or years forward.
Also, we are missing the value of the multiplier, a number which I'm sure comes with a pile of assumptions.

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep