Data in everyday life: Gas prices

Beth SavageOut of control? Gas prices are in the news, with nearly everyone watching their ups and downs. Looking at gas price data statistically, it’s easy to see whether or not they represent a stable system. What’s your guess?

Gas Prices Are Plunging
Gas Prices Lowest in 6 Years
How Low Will Gas Prices Go?

The headlines abound. Nearly every media outlet from the local small town daily news to the Wall Street Journal has a news story about gas prices on a weekly basis. It seems it’s news when they rise and it’s news when they fall. And every week Americans feel the impact when they fill their tanks—trying to time their fill-ups when prices are lowest.

While the gas prices are bouncing up and down like a Superball, I wondered if the prices are statistically different, or is the variation we are experiencing actually normal? To answer this question, I look to the control chart—the best tool for analyzing data to determine if a process is experiencing the normal variation that is inherent in every process, or if special cause variation is present.

Looking at the average weekly gas prices1 per quarter since 2005 reveals a process that is highly unstable. In fact, at least four types of out-of-control conditions exist:

  1. Points above the upper control limit
  2. Point below the lower control limit
  3. Runs below the mean
  4. A run above the mean

Flashback to the fourth quarter of 2008 (a painful place to go, I know.) The price of gas continued to go up and up, but due to the economic downturn, demand for gasoline was going down for the first time in years. With demand falling, and the dollar starting to make a comeback after the bank bailouts, the price of a barrel of crude oil dropped $10 in just one day. That sparked a major sell off of oil commodities, and burst the oil bubble.  As a result, the price of gasoline fell from a peak of nearly $4 a gallon to below $2 a gallon by the first quarter of 2009.

The first quarter of 2016 is trending below the lower control limit, and although prices are rising slightly, it won’t be enough to pull it back into normal range.

Now if I take a look at a control chart of the average weekly gas price each year (vs. each quarter) for the same time period, I find that the process is statistically stable. The weekly average price has been as high as $3.61 per gallon in 2011 and as low as $2.00 a gallon last year, but the process is actually quite stable.

We know that since 2005, weeks have existed where the average price per gallon has risen above the upper control limit of $4.19 and below the lower control limit of $1.68, but prices haven’t remained there for long. Thus “data point mentality,” or looking at data on a point to point basis and reacting to changes, makes newsworthy headlines, but it can also cause whiplash. Investigate data over time in ways that are meaningful. Identify special causes that require attention, and beware of overcontrolling, or reacting to variation that is a normal part of your process. Unless of course, you want to make headlines.

1Data source for gas prices: https://www.eia.gov/dnav/pet/pet_pri_gnd_dcus_nus_w.htm

2 thoughts on “Data in everyday life: Gas prices

  1. I never look at a control chart without first looking at the associated range chart. I think you are trying to present a point of view (gas prices are in control) using only a part of the data. Statistics never lie – but watch out for statisticians.

  2. Great point about the range charts, Jim. I agree that variability charts should be paired with charts for averages. In my zest to get the reader to focus on a predictable process in terms of the average, I ignored the variability chart.

Comments are closed.