Saturday, January 16, 2016

Chain-Weighting and Investment Ratios

A helpful reader of my last post alerted me to one of the pitfalls of chain-weighted national income accounts data, complete with references (this one and this one). Chain weighting, in principle, gives us better measures of real GDP and inflation, but it has some negative side effects. For example, the chain-weighted expenditure components (consumption, investment, etc.) of chain-weighted real GDP do not add up to chain-weighted GDP. Further, in a long time series, calculations of ratios of expenditure components can be thrown off significantly. In my last post, in fact, the investment ratios I was looking at were ratios of chain-weighted real investment to chain-weighted real GDP - my bad.

So, suppose we do this correctly, and calculate investment ratios as nominal investment to nominal GDP. If we do that, we get a ratio of gross private investment to GDP that looks like this:
So, investment is not booming, but the current investment rate, at 16.8%, is only slightly below the sample average (post-1947) of 17.2%, and the post-1980 average of 17.5%. Further, the investment measure in the chart includes residential investment, and maybe we think there are a set of special factors that apply to housing, for the issue at hand. So, if we do the calculation (again ratios of nominal expenditure components) for gross private nonresidential investment to GDP, we get
Again, that's not booming, but the current nonresidential investment rate is about at the post-1980 average, and is higher than the sample average of 12%.

So, my story is essentially the same, and is consistent with Rachel and Smith's measurements (though not with their story). The current investment rate is about at the post-1980 average, even if you include residential investment, but the real rate of return on capital is not low. So, we don't want to use a savings glut/secular stagnation story to explain why the real rate of return on government debt is low.

Addendum: Now I'm not sure this is better. But, apparently (I don't have references, but I have heard this said), the relative price of capital goods has fallen over time. So, wouldn't I want to use ratios of real investment to real GDP to measure the investment rate? Maybe chain weighting messes with that measure, but why is it any worse than the ones above, that don't account for the relative price change?


  1. I don't get why we're measuring investment as a ratio of nominal output at all, since there's clearly an endogenous relationship there? Wouldn't it be better to measure something like inflation adjusted investment per capita?

    1. "inflation adjusted investment per capita"

      which would have a trend in it.

  2. Not 100% sure this makes sense, but I think one way to address the issue in your addendum is to look at contributions to growth in real GDP (NIPA 1.1.2 or 3 I believe), which do sum to 1 and reflect relative price changes.