What the heck? Here’s what the Trump-Hassett CEA reports in February.
Source: ERP, 2019, as edited by MDC (orange arrow).
Consider carefully the bars highlighted by the orange arrow. If you look at the data for equipment investment in real terms, 2009Q3-2019Q1 (2nd release), one obtains the following graph. It’s hard to see how the pre-TCJA trend growth rate is negative.
Figure 1: Equipment investment, in bn Ch.2012$, SAAR (blue, log scale), stochastic trend estimated on logged data, 2009Q3-2017Q3 (red), +/- 1 se prediction interval (gray). TCJA period for enhanced expensing shaded orange. Source: BEA 2019Q1 second release, author’s calculations.
My calculations indicate that (stochastic) trend annual growth rate over the pre-TCJA period indicated by the ERP is 7.8%, not -2.2% as indicated in ERP Figure 1-6. The TCJA period growth rate is 6.4%, rather than the 6.6% reported. Using the vintage the CEA probably used (December 2018 release) does not change these fundamental patterns. It might be that CEA is using net rather than gross investment; if so, I could not find an indication that this was the case in the text. (Data here).
Interestingly, the most recent data shows the investment boom is faltering, and faltering significantly relative to pre-TCJA trend. (I use a stochastic trend because log equipment investment fails to reject the unit root null using ADF, and ERS-DF, and rejects trendstationary null using KPSS).
So, I’d welcome any thoughts on this puzzle — have I misread something or is there a mistake in the ERP?
Update,6/4 2PM Pacific:
Upon reflection, I think I have determined what is going on — and my confusion is partly due to my mis-reading. The key is to read the notes to the figure:
The overall rates for the nonresidential investment trend and actual compound annual growth are calculated based on a weighted average of the structures, equipment, and intellectual property product components.
This means the red bar is an actual growth rate calculated in the standard manner. The blue bar is the year-by-year decline in the trend growth rate in investment — not the corresponding average annual growth rate over the pre-TCJA period. I think the way this is number is calculated is by estimating a quadratic trend, and calculating the implied annualized decline in the growth rate. I get pretty close to the figures in the bar charts. The difference might be attributable to how the CEA staff estimated the pre-TCJA trend (deterministic or stochastic; see this post for discussion).
I will say that the notes to this figure are not the most transparent; indeed, one might say they are misleading, given that the vertical axis is actually denoting two different variables even though the text suggests the same variable over two periods is being plotted.
Update, 6/6 8:45AM Pacific:
Here is my re-interpretation of CEA Figure 1-6.
Figure 2: Chinn’s reinterpretation of ERP, 2019 Figure 1-6. Trend growth rates calculated using log differences. Nonresidential investment based on reported aggregates. Equipment investment calculations uses data up to 2017Q3 for the pre-TCJA period. Source: BEA 2018Q3 final release, and author’s calculations.