Junk Charts: Secondary Axes

by Stubborn Mule on 6 October 2009 · 8 comments

Data visualisation is a fascination of mine and, as a result, the posts on this blog have featured a wide array of charts. My aim is always to use graphics to help explore data and provide greater insight into whatever phenomena might be lurking in the data. While I have tried to make use of “good” charts, I have never talked about “bad” charts. There is no end to the things that can be said about bad charts. Indeed, one of my favourite blogs, Junk Charts, is dedicated to the subject. Nevertheless, I have decided that I will begin to assemble some bad charting habits to avoid.

Inspired by a chart on the Australian Stock Exchange (ASX) web site, brought to my attention by Danny Yee, I will start by considering the perils of the “secondary axis”. This is a topic that has been covered on Junk Charts, but it seems to be poorly understood by many people in finance. Bank economists seem to be particularly fond of secondary axes.

Since the ASX chart is dynamically generated, I have reproduced a static snapshot showing the key features of the chart. The chart shows (in blue) the history of the S&P/ASX 200 share price index over the last six months. The axis on the left hand side is the “primary axis” and shows the values of this share price index. Overlaid in red is the share price of Alchemia Limited (ACL), a biotech company. The price of Alchemia is shown on the axis on the right hand side, the “secondary axis”.

ACL - ASX version

Alchemia Share Price on a secondary axis (on the right)

At face value, it would appear that the price of Alchemia shares has tracked the performance of the overall share market very closely. However, this is an illusion created by the use of a secondary axis. When considering performance of shares, what is important is not the share price itself, but the returns these prices generate. The upper and lower limits of each axis bear no relation to one another, but are simply determined by the range of values each price series takes and so give no insight into the returns of Alchemia compared to those of the S&P/ASX 200.

A better approach is to create a return index by considering what would happen to $100 invested in either Alchemia or the market as a whole (ignoring dividends and any costs such as brokerage fees and stamp duty). Both these indices can then be plotted on the same axis and this gives a very different picture.

ACL - relative

Alchemia Share Price on a common axis

It now becomes clear that Alchemia shares performed dramatically better than the market as a whole. In fact, over this six month period, while the sharemarket index  returned a very healthy 32%, the return on the Alchemia share price was an enormous 304%. This return was helped, no doubt, by Alchemia’s progress in drug trials with the US FDA.

The most that can be said for the ASX chart with the secondary axis is that it reveals some correlation in the ups and downs of the Alchemia share price and the broader market. However, the chart completely missed the big story in the data.

It is worth noting that producing the secondary axis in the first chart above using the R package is rather fiddly, while creating secondary axes in Microsoft Excel is very straightforward. This points to what I expect to be a theme in future bad chart posts: if it is easy to do in Excel, it is probably bad.

Alchemia

Possibly Related Posts (automatically generated):

{ 6 comments… read them below or add one }

1 Stilgherrian October 6, 2009 at 3:25 pm

Your slogan “If it it’s easy to do in Excel, it is probably bad” also works if you replace “Excel” with “Las Vegas” or “Patpong Road, Bangkok”.

2 stubbornmule October 6, 2009 at 3:28 pm

You’ve got a point there Stil. Perhaps I should be more specific: “it’s probably bad, but at least you shouldn’t catch anything”.

3 Danny Yee October 6, 2009 at 4:12 pm

I have a very low opinion of “technical analysis” (as an investment tool) at the best of times. But when done using charts like this – or charts of share prices that completely ignore massively dilutive capital raisings – it starts to make astrology seem empirical.

4 Jon Peltier October 7, 2009 at 11:55 am

“if it is easy to do in Excel, it is probably bad”

Secondary axes, pie charts, 3D charts, all that formatting. Yeah, that stuff is too easy to do, and bad.

5 Marc October 7, 2009 at 4:10 pm

Great article – this is a particular bug bear of mine.

I think you might like this: http://www.mrscienceshow.com/2008/06/economists-oil-cricket-and-correlation.html

With secondary axes and a bit of deviousness, you can may anything look related to anything – in this case, aus cricket team performance and the price of oil – here was my followup to that, showing why it was dodgy in the first place: http://www.mrscienceshow.com/2009/02/poor-correlations-or-why-its-not-fault.html

6 Sophia Bennett October 9, 2009 at 2:13 am

Hi Sean

I can’t believe I not only like but actually understand this post! Hooray for your powers of persuasion!

Leave a Comment

 

{ 2 trackbacks }

Previous post:

Next post: