I came across this chart on Twitter.
It claims that there is a 97% correlation but to get that tight fit they use two different scales on the primary and secondary y-axis.
The scale on the black axis is a 500% range from bottom to top and the scale on red axis is 271% from top to bottom. This means the creator of the chart believes a 5% move is equivalent to a 2.7% move.
Further they compress the x-axis too. In black it has 9/1982 to 6/1983 which is 9 months but then the corresponding red period is 4/2014 to 12/2014 is 8 months. But those time periods are aligned. The next period they adjust is 10/1984 to 7/1985 which is 9 months and it is aligned with 5/2016 to 1/2017 which is 7 months. Although the total interval lengths for red and black dates are the same length, the creators compressed the data to make it fit.
THIS IS NOT CORRELATION.
This is how the chart would look if you normalized each data series and started at 1.
And this is how the chart would look if you used the same scales for the y-axis.
Note on my two charts that when you dont alternate between 7 or 8 or 9 months to compress it the peaks and valleys do not align as perfect. Note how the the 1982-1987 market moved substantially higher than the 2014-2018 market did?
AGAIN THIS IS NOT CORRELATION. When you data mine and compress the data and use different scales you are not doing a correlation. A 5% move is not correlated with a 2.7% move at a 97% rate.
Investors beware of charts like this.
If you like ideas like this please sign up to the free quarterly economics newsletter here