From f90ed78ffc00bb145af3904658cc5ff0b1429646 Mon Sep 17 00:00:00 2001 From: Matt McKay Date: Tue, 18 Aug 2026 19:12:05 +1000 Subject: [PATCH 1/2] =?UTF-8?q?test=5Fpwt.csv.yml:=20record=20the=20true?= =?UTF-8?q?=20PWT=206.3=20overlap=20=E2=80=94=20four=20POP=20and=20four=20?= =?UTF-8?q?XRAT=20values,=20not=20two?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The header's vintage-elimination trail said PWT 6.3 "matches Argentina and Australia POP but nothing else". Re-measured against the same archive (member sha256 f9609c42…), year 2000, exact float equality: 6.3 matches four of eight POP values (ARG, AUS, ISR, ZAF) and four of eight XRAT values, with no cc or cg value and no tcgdp column. The overlap is the hint any future vintage hunt would follow, so understating it is what this corrects. The adjacent 6.2/6.1 "match nothing" gets the same treatment from the same measurement: no POP, cc or cg value, with XRAT agreeing only on ARG+USA (6.2) and USA (6.1). The central conclusion stands as written — the committed values match no downloadable PWT vintage — and the four archive hashes were all re-derived and confirmed. Comment-only change: the parsed record, CATALOG.md and the strict audit are untouched. Found during the QuantEcon/data-lectures#100 validation. Co-Authored-By: Claude Fable 5 --- lectures/test_pwt.csv.yml | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/lectures/test_pwt.csv.yml b/lectures/test_pwt.csv.yml index c3c5ce4..000bdd5 100644 --- a/lectures/test_pwt.csv.yml +++ b/lectures/test_pwt.csv.yml @@ -13,10 +13,11 @@ # against four archives from rug.nl/ggdc: PWT 7.0 (pwt70_06032011version.zip, # member pwt70_w_country_names.csv sha256 4623b92a…) has different tcgdp/cc/cg # for every row and renamed the isocode column; PWT 6.3 -# (pwt63_nov182009version.zip, member sha256 f9609c42…) matches Argentina and -# Australia POP but nothing else and has no tcgdp column; PWT 6.2 -# (pwt62_data.xlsx, sha256 a3293337…) and PWT 6.1 (pwt61_data.xlsx, sha256 -# 1477a118…) match nothing. The committed +# (pwt63_nov182009version.zip, member sha256 f9609c42…) matches four of eight +# POP values (ARG, AUS, ISR, ZAF) and four of eight XRAT values exactly, but +# no cc or cg value, and has no tcgdp column; PWT 6.2 (pwt62_data.xlsx, sha256 +# a3293337…) and PWT 6.1 (pwt61_data.xlsx, sha256 1477a118…) match no POP, cc +# or cg value (XRAT agrees only on ARG+USA in 6.2 and USA in 6.1). The committed # header uses the 6.x-era column name `country isocode` while `tcgdp` appears # only in the 7.x releases checked — so the extract mixes or predates the # vintages still downloadable. Hence `unrecovered`, with this paragraph as the From c5e0dd543ef6f632531fa60f3704cb205bbbf297 Mon Sep 17 00:00:00 2001 From: Matt McKay Date: Tue, 18 Aug 2026 19:19:40 +1000 Subject: [PATCH 2/2] Apply 'exact float equality' to both columns and enumerate the XRAT matches Copilot review: the trailing 'exactly' could read as qualifying only XRAT, and the XRAT matches were not enumerated the way the POP ones are. The qualifier moves to the front of the pair, and the four 6.3 XRAT matches (ARG, MWI, USA, URY) are now recorded parallel to POP. Co-Authored-By: Claude Fable 5 --- lectures/test_pwt.csv.yml | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/lectures/test_pwt.csv.yml b/lectures/test_pwt.csv.yml index 000bdd5..e4b4972 100644 --- a/lectures/test_pwt.csv.yml +++ b/lectures/test_pwt.csv.yml @@ -13,9 +13,10 @@ # against four archives from rug.nl/ggdc: PWT 7.0 (pwt70_06032011version.zip, # member pwt70_w_country_names.csv sha256 4623b92a…) has different tcgdp/cc/cg # for every row and renamed the isocode column; PWT 6.3 -# (pwt63_nov182009version.zip, member sha256 f9609c42…) matches four of eight -# POP values (ARG, AUS, ISR, ZAF) and four of eight XRAT values exactly, but -# no cc or cg value, and has no tcgdp column; PWT 6.2 (pwt62_data.xlsx, sha256 +# (pwt63_nov182009version.zip, member sha256 f9609c42…) matches, at exact +# float equality, four of eight POP values (ARG, AUS, ISR, ZAF) and four of +# eight XRAT values (ARG, MWI, USA, URY), but no cc or cg value, and has no +# tcgdp column; PWT 6.2 (pwt62_data.xlsx, sha256 # a3293337…) and PWT 6.1 (pwt61_data.xlsx, sha256 1477a118…) match no POP, cc # or cg value (XRAT agrees only on ARG+USA in 6.2 and USA in 6.1). The committed # header uses the 6.x-era column name `country isocode` while `tcgdp` appears