bugfix in Penn World Table unit conversion

[orichters]

• REMIND shows unintended capital / consumption jumps between 2005 and 2015, see this remindmodel issue
• analyzing AMT runs, it was found this changed between November and beginning of December while switching from cfg$inputRevision <- "6.316" to "6.324" as visible in this comparison
• Michaja found that calcCapital() results differ between 1242136 and 298465a, see this mrremind issue
• I think responsible is this PR in mrdrivers which adds to calcInternalGDPPastPWT() the following line:

data <- GDPuc::convertGDP(data, "constant 2017 Int$PPP", unit, replace_NAs = c("linear", "no_conversion"))

PENN World Table data is read from this file called /p/projects/rd3mod/inputdata/sources/PWT/pwt80.xlsx, which states:

rgdpna: Real GDP at constant 2005 national prices (in mil. 2005US$)

I think this data must not be converted from 2017 to 2005 data, and the PR description does not provide a rationale why this change was intended. Therefore, I suggest to revert this change.

[orichters]

No idea why buildLibrary deleted this file…

[0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q]

Anybody made new input data to

1. confirm this fixes the bug
2. confirm this does not introduce "unwanted behaviour" somewhere else?

[orichters]

Lavinia started input generation yesterday and we will check the results.

[0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q]

Looks good.

$ tar -Oxf /p/projects/rd3mod/inputdata/output/rev6.47_62eff8f7_remind.tgz ./f29_capitalQuantity.cs4r | grep "20[01].,EUR,gdp_SSP2EU,kap,"
2000,EUR,gdp_SSP2EU,kap,37.824154
2005,EUR,gdp_SSP2EU,kap,45.493689
2010,EUR,gdp_SSP2EU,kap,47.917582
2015,EUR,gdp_SSP2EU,kap,50.561141

[mleimbach]

Oli, can you double check. For REF I see these input numbers, which are higher than what we have before:
tar -Oxf /p/projects/rd3mod/inputdata/output/rev6.47_62eff8f7_remind.tgz ./f29_capitalQuantity.cs4r | grep "20[01].,REF,gdp_SSP2EU,kap,"
2000,REF,gdp_SSP2EU,kap,10.074331
2005,REF,gdp_SSP2EU,kap,9.44623
2010,REF,gdp_SSP2EU,kap,11.412844
2015,REF,gdp_SSP2EU,kap,12.458628

[orichters]

Well, this was the change from 6.316 to 6.324, reflecting the difference between Nov-26 and Dec-03 AMT runs:

> tar -Oxf /p/projects/rd3mod/inputdata/output/rev6.316_62eff8f7_remind.tgz ./f29_capitalQuantity.cs4r | grep "20[01].,REF,gdp_SSP2EU,kap,"
2000,REF,gdp_SSP2EU,kap,9.997588
2005,REF,gdp_SSP2EU,kap,9.382936
2010,REF,gdp_SSP2EU,kap,11.336092
2015,REF,gdp_SSP2EU,kap,12.446041

> tar -Oxf /p/projects/rd3mod/inputdata/output/rev6.324_62eff8f7_remind.tgz ./f29_capitalQuantity.cs4r | grep "20[01].,REF,gdp_SSP2EU,kap,"
2000,REF,gdp_SSP2EU,kap,15.941333
2005,REF,gdp_SSP2EU,kap,14.905243
2010,REF,gdp_SSP2EU,kap,17.731448
2015,REF,gdp_SSP2EU,kap,19.179349

So we are basically back to the old results, which I think is fine. I don't understand where the small difference to 6.47 comes from, but the difference is less than 1%.

[0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q]

[mleimbach]

How are these numbers related to what I would get when I plot from the fulldata.gdx (which also correpond to your plots in the macro group chat):

1 SSP2EU-AMT-Base_2022-11-26 REF 2005 3.54 kap quantity
2 SSP2EU-AMT-Base_2022-12-03 REF 2005 5.72 kap quantity
3 SSP2EU-AMT-Base_2022-11-26 REF 2010 4.27 kap quantity
4 SSP2EU-AMT-Base_2022-12-03 REF 2010 6.80 kap quantity
5 SSP2EU-AMT-Base_2022-11-26 REF 2050 11.6 kap quantity
6 SSP2EU-AMT-Base_2022-12-03 REF 2050 17.4 kap quantity

[orichters]

My suspicion: MER vs PPP, done in this conversion here, where the capital factor from the input data is multiplied with 0.384747 for REF to get the data used by REMIND. And 9.382936 * 0.384747 = 3,61006, and then some specific capital stocks for industry etc. are subtracted, yielding the 3.54 value.

[mleimbach]

Thanks Oli. Right, input data are in PPP and REMIND data in MER...
It's just not to overlook something.