If you don’t believe me, please take a look at the time series chart of US GDP per capita available.
figure 1: GDP per capita, in Ch.2012 $ SAAR (blue); linear trend (brown). Source: BEA through FRED, and the author’s calculations.
Using robust standard errors, linear regression produces a t-stat over a time trend of approximately 60. The GDP per capita increases by about 40 Ch2012$ every quarter.
Or, one might want to plot on a logarithmic scale.
figure 2: GDP per capita, in Ch.2012 $ SAAR (blue, logarithmic scale). Source: BEA through FRED, and the author’s calculations.
Which one is misleading?one reader think:
Generally speaking, all audiences except professional audiences should avoid using logarithmic scales, especially when it comes to exponential growth rates. I rarely use them in business environments or general audiences because they can be easily misunderstood. If possible, I really like simple linear scales.
Well, I think people can better describe this series as exhibiting exponential growth (unit root) rather than unit growth.




