Is GDP accurate? Barro argues that our measures of GDP are systematically biased because they include investment both when it occurs, and when we get higher output as a result. Thus, higher capital/output places will have spuriously higher GDP.

First, though, we should know what GDP is trying to measure. It is not measuring welfare — for that would take knowledge of how pleasant work is — but simply trying to give a sense of the resources available for consumption. This means that we don’t count intermediate goods. GDP is computed as net value-added, or else GDP would be arbitrarily increased by the number of steps in the production process. The wheels, chassis, and body of a truck, for example, is only valuable insofar as it’s a truck. Let’s take the example of a truck and work with it. Once we’ve made it, a truck is often used to produce other things. It can carry materials from place to place. This should properly make it an intermediate good. The BEA, however, treats a truck produced as a final good. Suppose there are two countries A and B, with A having a higher level of technology than B. B, however, has more capital. We can imagine a world where both places have the exact same level of income, but B has a higher GDP. The trucks which are used in country B count once when they are made, and once again when they show up in higher income per person. Thus, having more capital overstates the actual income of a country. This is obviously imperfect. We need to be able to tell what percentage of an investment good is itself consumption, and how much investment. (Take a pick-up truck. It takes you to work, but is also nice to drive. How much of it is investment?) This is likely better, though, then arbitrarily increasing or decreasing GDP whenever we decide to include a different form of capital as an investment good or not.

Robert Barro, “Double-Counting Investment” (2020) https://scholar.harvard.edu/files/barro/files/double-counting_paper_jan_2020.pdf