What source(s) do we use to derive industry (x) and commodity (q) output for the IO model year and other model years?

[WesIngwersen]

Industry output. $x$ and commodity output, $q$, vectors, for the year of the IOT and for other years present can be derived from multiple sources.

USEEIO

For IOT year, USEEIO derives from the Use table, where $x$ is the column sums of industry columns, and $q$ is the row sums of all commodity rows. Reference code

For non-IOT years, for $x$, USEEIO uses the BEA annual gross output (UGdpByInd) data after being mapped to the IO sectors.
Reference code.
To estimate $q$ for non-IOT years, USEEIO calculates the commodity mix matrix from the Make table from the IOT year and pre-multiplies the commodity mix matrix by an industry output matrix for every year.

[MoLi7]

CEDA

For IOT year (benchmark year, e.g. 2017), both $x$ and $q$ are derived from the Make table, where $x$ is the row sums of industry rows, and $q$ is the column sums of commodity columns. Reference code here and here.

For non-IOT years, CEDA first scales $q$ using the Summary $q$ vectors from annual IOTs from original_year (2017 in this case) to target_year (the latest year available in annual IOTs); then if target_year is still before ceda_base_year, CEDA adjust $q$ again via inflation adjustment, i.e. using chain-type price index. Note if target_year == ceda_base_year, the inflation adjustment step would have no effect on $q$.
$x$ is not estimated for non-IOT years in CEDA.

[bl-young]

Maybe two separately related questions.

1. Do we derive q/x from the Make table or the Use table. For a benchmark year, given balanced tables that are not mixed unit or hybridized or disaggregated this should be the same? We should be able to calculate both and confirm they are the same. (e.g. example of adjustment made to do this for an MUIO)

2. How do we handle deriving these for non benchmark years where commodity output is not provided by BEA. Needs further research and evaluation of two approaches