> Why is US GDP growth so weirdly constant? i'm not sure about this but i get the impression that semilog plots of exponentially growing things often look dece

> Why is US GDP growth so weirdly constant? i'm not sure about this but i get the impression that semilog plots of exponentially growing things often look deceptively linear, and hence that the best-fit exponential curve often looks like a better fit than it is. The claim here is that the actual rate-of-growth has been constant over time. What is shown is a graph of GDP growth with an exponential curve overlayed whose rate-of-growth parameter has been fit to the data. But perhaps if you had stopped the analysis at different times in the past, you would have gotten similarly linear-looking graphs, but with very different best-fit rate-of-growth parameters. What i might like to see instead is a graph showing, for each year in the past, the value of the best-fit rate-of-growth parameter if you looked only at a fixed-sized moving window of data ending at that year. If the size of the moving window is '1 year', it is no surprise that GDP is all over the map. But even when the window is 10 years, charts of GDP growth per decade (from a google image search) show more substantial variation: http://www.leftbusinessobserver.com/GDP-per-cap-by-dec.jpg http://www.leftbusinessobserver.com/GDP-per-cap-by-dec.jpg http://www.massline.org/Dictionary/Photos/G/GDP-US-CumulativeGrowthByDecade.jpg http://www.massline.org/Dictionary/Photos/G/GDP-US-Cumulativ... I expect that the variation would be even more pronounced if you made a chart that showed, for each year, the growth over the preceding 10-year window, instead of those bar charts which only have one bar every 10 years. Also, even 10 years is too short; what i'd really like to see is charts of average GDP growth over 30 years, but i don't have time to compute that. In addition, there is some evidence that very high rates of GDP growth are only possible when a country is less wealthy/developed (see eg figure 4.4 from https://www.econ.nyu.edu/user/debraj/Courses/GrDev17Warwick/Notes/RayCh3UpdatePart2.pdf https://www.econ.nyu.edu/user/debraj/Courses/GrDev17Warwick/... ; note that it appears to be possible for less wealthy/developed countries to have either low or high GDP growth, whereas wealthy/developed countries appear to only be capable of low GDP growth; this could be explained by a 'diminishing returns' argument). If true, this would imply that over very long timescales, the rate of growth would decrease. My "analysis" here is rather hacky -- what is the proper statistical way to analyze this? What is the proper statistical test to see if long-term growth rates are constant?