As several commenters in the past have questioned Dean’s estimates (which I am convinced are robust, especially through March 2016), I asked Dean the following questions:
“At one point you thought there may have been a structural change in the data. Is that no longer the case? …. Are you seeing any issues that I … would have missed?”
He answered:
“I just tested the correcting factors for stationarity with panel unit root tests and the null of a unit root is rejected for all tests considered.”
I did some research on stationarity and unit root tests and then asked:
“So the mean and variance of the correcting factors do not change with time or follow any trend?”
He said that was correct.
Essentially the statistics suggest that the mean and variance of the correction factors have remained stable over the past 24 months and over time they are likely to approach their “true” value. The Chart below shows how the TX C+C correction factors have changed from Sept 2014 to April 2016.
The chart below shows how the estimates of TX C+C have changed from May 2015 to April 2016. The dashed lines suggest the estimates with the minimum correction factors (Nov 2015) and the maximum correction factors (March 2016).
…click on the above link to read the rest of the article…
