Gordon's Paper has been thoroughly investigated by Roger Pielke Jnr at the Breakthrough Institute.
Gordon's smoothing of growth fails to show the variability and creates a picture of trends that are not really there. A quote from the article linked above:
In short, there is no evidence of a stair step reduction in the growth rate of US per capita GDP in either dataset. The US BEA and Census data shows essentially no change (a linear trend, blue line, shows a statistically insignificant downward tick) whereas the Maddison data shows a bit of an increase (red line). The data is sensitive to the time period chosen – for instance, from 1970 the BEA/Census data shows an increase in the annual rate of per capita GDP growth. I can find no evidence of a post-1950 secular decline in per capita economic growth in the United States, and in fact, there is evidence that growth rates have accelerated a bit from 1970.