4
$\begingroup$

I found the axis on the following graph of crude oil prices over 1950-2015 from MacroTrends surprising. enter image description here The y-axis labels are 20, 40, 60, 80, 100, ...but they are not equidistant. Why?

$\endgroup$
6
$\begingroup$

Usually, when people use non-equidistant axis, it is because they want to emphasize some variation more than others. For example, if most of your data is between 0 and 1, but you have one outlier on 100, then an equidistant axis would make it hard to analyze most the variation, because it emphasizes the wrong parts.

In this case you have a log-scale, which is a simple nonlinear axis that (in general) emphasizes small values over larger ones. Whether that makes sense with oil prices, is - to me - unclear.

$\endgroup$
  • 4
    $\begingroup$ It might be worth pointing out that specifically, a log scale emphasizes volatility (b/c equal % changes show up as equal movements, rather than equal absolute changes having equal movements), which can definitely be of interest with oil prices. Also, a log scale can be very appropriate for series (like GDP level) that are characterized by compound growth. The use of both a log scale and an inflation adjustment (inflation-adjusting oil prices complicates many analyses, due to the powerful endogeneity) is pretty hard to justify, though. $\endgroup$ – dismalscience May 6 '15 at 20:44
  • $\begingroup$ @dismalscience, you should write this up as an alternative answer. It is often good to graph transformed data to present a stationary time series. This is especially important when time series are long. While it is true that as foobar says, this isn't very important for oil prices because the technology of extraction and refining have improved almost as fast at nominal prices have increased, it is true for most long run price series. Another alternative is to graph the levels but to deflate the series to show real (rather than nominal) prices. $\endgroup$ – BKay May 7 '15 at 11:49
3
$\begingroup$

As it has already been explained, it's a logarithmic scale (also called log scale for short). The distance between values doesn't depend on their linear difference, but on their relative difference. The distance between 20 and 40 is the same as the distance between 40 and 80 or the distance between 60 and 120, because 40/20 = 80/40 = 120/60 = 2 or a 100% increase, a fixed rate.

An additional caveat about log scales is that they never go down to 0. After all, if you increase 0 by any percentage, you still get 0. The distance between 0 and any positive value is infinite.

$\endgroup$
  • $\begingroup$ Your caveat is negligible when plotting prices, though. $\endgroup$ – FooBar May 6 '15 at 17:54
  • $\begingroup$ Yeah. I was thinking about log scales for any kind of data, not necessarily prices. My comment also referred to the André Peseur's last paragraph saying that the Y axis was heavily truncated. If "truncated" means that it doesn't go down to 0, well, log scales just need to be truncated at some point - maybe the lowest value in the data set or some value not much lower than that. $\endgroup$ – dgstranz May 6 '15 at 18:27
  • $\begingroup$ Of course, you have to stop a point point. But looking at the graph it seems that they have stopped pretty early. $\endgroup$ – user4239 May 6 '15 at 19:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.