super informative article, thank you! I am trying to get to the 665B annlualized number but not sure how. I see wages and salaries in the Q2 GDP report for 11,108.4 (in billions). So do I need to come up with 10,441 (in billions) from CES establishment data somehow?

Hi Zeno! Glad you enjoyed the article. Conceptually you can treat the GDI private sector wages data as approximately equivalent to average hourly earnings*private nonfarm employment*average weekly hours, and if you plot that (indexed to 2019) the gap that emerged was about 6% of the GDI wages and salaries number, which is how I got the $665B annualized number.

Revisions just occurred to the GDI data today, and the BEA's wages numbers have been revised downwards—so if you are looking at the gap right now, it will be much smaller than it was when I wrote this post.

Suppose we had a firm with $100 in sales, and it reported $80 in expenses and $40 in profits. There would be a statistical discrepancy of $20 by which income exceeds sales.

It might be the case that the firm's measure of expenses includes $15 in bonuses that are not included in somebody else's measure of the firm's expenses. That is interesting, but I don't see how it explains the statistical discrepancy. Unless you are saying that the $15 in bonuses is exaggerated and that it will be revised away, so that expenses will turn out to be $65 and the statistical discrepancy will be reduced that way.

A better analogy would be something like this: imagine you run a flower shop that sells $100 of flowers. You pay your workers $85 in wages and take $10 in profits. To keep workers from leaving, you decide to give them a new, special $5 bonus at the end of the month.

When calculating GDP data, the BEA will look and say "well, $100 of flowers were sold here so GDP is $100." Then when calculating GDI they will go "well, workers were paid $85 and the owners took $10 in profit, so GDI is $95." So there's a $5 gap between the GDP and GDI estimates that doesn't close until GDI is revised to include the bonus!

Right, but my point is just that BEA just extrapolates from last year using CES data for its initial estimates and then accounts for changes in bonuses and the like when QCEW data is incorporated. If you had given a bonus in 2021 but not 2022, BEA would be overestimating when it extrapolated with CES data and then would revise down when it incorporated QCEW data.

sorry, trying to follow here. BEA's third revision of GDP showed that wages and salaries were revised down. how does this affect the CES vs QCEW debate? maybe i am confusing all kinds of numbers here but if you could shine some light onto this, it would be very helpful.

Amazingly comprehensive analysis of the discrepancies between GDP and GDI and thanks for your hard work!

Thank you Robert!!

Hi Joseph,

super informative article, thank you! I am trying to get to the 665B annlualized number but not sure how. I see wages and salaries in the Q2 GDP report for 11,108.4 (in billions). So do I need to come up with 10,441 (in billions) from CES establishment data somehow?

Hi Zeno! Glad you enjoyed the article. Conceptually you can treat the GDI private sector wages data as approximately equivalent to average hourly earnings*private nonfarm employment*average weekly hours, and if you plot that (indexed to 2019) the gap that emerged was about 6% of the GDI wages and salaries number, which is how I got the $665B annualized number.

Revisions just occurred to the GDI data today, and the BEA's wages numbers have been revised downwards—so if you are looking at the gap right now, it will be much smaller than it was when I wrote this post.

Appreciate the response, will try that.

Help me understand.

Suppose we had a firm with $100 in sales, and it reported $80 in expenses and $40 in profits. There would be a statistical discrepancy of $20 by which income exceeds sales.

It might be the case that the firm's measure of expenses includes $15 in bonuses that are not included in somebody else's measure of the firm's expenses. That is interesting, but I don't see how it explains the statistical discrepancy. Unless you are saying that the $15 in bonuses is exaggerated and that it will be revised away, so that expenses will turn out to be $65 and the statistical discrepancy will be reduced that way.

Hi Arnold—sorry for the confusion!

A better analogy would be something like this: imagine you run a flower shop that sells $100 of flowers. You pay your workers $85 in wages and take $10 in profits. To keep workers from leaving, you decide to give them a new, special $5 bonus at the end of the month.

When calculating GDP data, the BEA will look and say "well, $100 of flowers were sold here so GDP is $100." Then when calculating GDI they will go "well, workers were paid $85 and the owners took $10 in profit, so GDI is $95." So there's a $5 gap between the GDP and GDI estimates that doesn't close until GDI is revised to include the bonus!

But I thought GDI what is greater than GDP￼

Right, but my point is just that BEA just extrapolates from last year using CES data for its initial estimates and then accounts for changes in bonuses and the like when QCEW data is incorporated. If you had given a bonus in 2021 but not 2022, BEA would be overestimating when it extrapolated with CES data and then would revise down when it incorporated QCEW data.

sorry, trying to follow here. BEA's third revision of GDP showed that wages and salaries were revised down. how does this affect the CES vs QCEW debate? maybe i am confusing all kinds of numbers here but if you could shine some light onto this, it would be very helpful.

So GDI now is counting more bonus income than actually was paid so far in 2022

It could be! We don't have enough evidence to say that for certain, and won't know until some more revisions to GDI data in late 2022

I would bet on an upward revision to GDP. The drop in productivity implied by current estimates looks suspicious