Pre-2013 and Post 2013 Comparisons

The attached link is a short document from the Census Bureau itself addressing the changes which they made in 2014.

It seems that you are generally right that income is not directly comparable between pre-2013 and post-2013. However, other results (such as earnings) are comparable.

When you compare earnings pre-2013 and post-2013 (see below), the earnings have gone up. However, there was not a statistically significant difference in earnings between 2015 and 2016.

https://www.census.gov/content/dam/Census/library/visualizations/2017/demo/p60-259/figure3.pdf

As with most data, numbers are subject to different interpretations but this should give more nuance than "the methodology changed...its all garbage" argument.

Post 2013 Comparisons

My biggest problem with your argument (and ZeroHedge link) is that we now have four years of data using the new information (2013, 2014, 2015, 2016). Any change in methodology that has been constant for all four years should bias all of these years equally. The results (from the first Zero Hedge graph) are a slight increase in 2013, a medium decrease in 2014, a huge gain in 2015, and a huge gain in 2016.

So while the whole dataset may be skewed based on the 2013 changes (and I think you exaggerate this effect), it is without a doubt that 2015 and 2016 were much, much better using comparable data.

Lastly, the oil argument is strange to me. The ZeroHedge article is arguing for changes year-over-year. So even if they are right that the oil prices being lower in 2015 artificially skewed the 2015 results vis-a-vis the 2014 results, oil prices were generally the same in 2015 and 2016. Thus lower oil prices do not account for any of the yearly change between 2015 and 2016 which is the data currently being discussed.

I know this was a long post, but I think more nuance gives us more faith in those numbers, not less. Just pointing out methodology change does not tell us much.