"Final GDP"

I remember reading long ago that "Banks have doubled their share of GDP to 8% from 4%".

I read it again recently in a PDF by Robin Greenwood and David Scharfstein:

During the last 30 years, the financial services sector has grown enormously. This growth is apparent whether one measures the financial sector by its share of GDP, by the quantity of financial assets, by employment, or by average wages. At its peak in 2006, the financial services sector contributed 8.3 percent to US GDP, compared to 4.9 percent in 1980 and 2.8 percent in 1950.

I can also find increase to 8% at FRED:

Graph #1: GVA of Finance as a Percent of GDP

But 8% never seemed quite right, somehow. The number just isn't big enough. All by itself, Monetary Interest Paid has never been as low as 8%, not once since 1960. It was 8½% in 1960:

Graph #2: Monetary Interest Paid as a Percent of GDP

In recent years, with interest rates low as they can go, interest payments are between 13 and 14 percent of GDP. And that's just the interest. How could finance be only 8%?

I wasn't the only one thinking such thoughts. Here is Jacob Assa in an article at Developing Economics, from January of last year:

Recent research has suggested that the imputation of productive output for the FIRE sector in the national accounts has distorted GDP in several ways. First, the FIRE imputation drives a wedge between output (as measured by GDP and employment trends), as visible in the phenomenon of ‘jobless growth’ observed after the three most recent recessions in the US (see figure 2).

Figure 2: US GDP (constant prices) and total employment, 1977=100

Second, real per capita GDP has likewise diverged from a more realistic measure of standard of living, real median income...

These and other anomalies have been investigated by recent research, which has also proposed some corrected measures of output. Basu and Foley (2013), in their paper Dynamics of output and employment in the US economy, exclude the FIRE sector from their measure of output (called Non-financial Value Added or NFVA) and find a much improved correlation between NFVA and GDP. In my own research, I go a step further, first excluding and then deducting financial fees from GDP. This is symmetrical to the treatment of financial interest income (FISIM) in the SNA, and also echoes Nordhaus and Tobin’s treatment of some government services as ‘instrumental’, i.e. intermediate consumption. I call this adjusted measure Final GDP (FGDP), since it treats finance as an intermediate input to the rest of the (productive) economy, and deducts it as such (for a full discussion of this methodology and its implications for economic theory and policy, see my new book on the Financialization of GDP).
Figure 4 shows GDP against FGDP and NFGDP (the GDP equivalent of Basu and Foley’s NFVA). Both alternative measures match employment better than GDP, with FGDP being the closest. FGDP also fits median income far better than per capita GDP, and even performs better than GDP in forecasting employment trends (see this working paper).

Figure 4: Three measures of US output (constant prices) and employment, 1977=100

[end of excerpt]

On the last graph above, the blue line shows GDP, which considers finance as "output" and includes it in the total. The red line omits finance. The green line treats finance as an intermediate cost and subtracts it from the total. To me the subtraction makes good sense.

Part 3 of 4