Three tests, none of which needs a source outside the file. Indonesian minimum wages are nominally non-decreasing, and in this file no province has a 2020 wage above its 2022 wage — so a 2021 value outside that bracket is not surprising, it is impossible.

test one · which year disagrees with its neighbours

correlation between two wage columns	ρ
2020 × 2022 · the two that skip it	+0.9958
2020 × 2021	+0.6812
2021 × 2022	+0.6723
2019 × 2020 · a normal pair, for scale	+0.9944

The two years either side of it agree almost perfectly with each other while both disagree with 2021. A real year cannot do that: to sit between two years that match, it would have to match them too.

test two and three · which rows, and whose numbers

province	2020	in file	2022	the value belongs to	off by	row gap
JAMBI	2,630,162	2,788,826	2,698,941	GORONTALO 2020	exact	+24
BENGKULU	2,213,604	2,460,996	2,238,094	BANTEN 2020	Rp 1	+9
KEP. BANGKA BELITUNG	3,230,024	3,000,804	3,264,884	KALIMANTAN UTARA 2020	exact	+15
DKI JAKARTA	4,276,350	1,765,000	4,641,854	— matches nothing in the file
JAWA BARAT	1,810,351	2,630,162	1,841,487	JAMBI 2020	exact	−7
DI YOGYAKARTA	1,704,608	2,215,000	1,840,916	— matches nothing in the file
BANTEN	2,460,997	3,230,023	2,501,203	KEP. BANGKA BELITUNG 2020	Rp 1	−7
KALIMANTAN BARAT	2,399,699	1,868,777	2,434,328	— matches nothing in the file
KALIMANTAN TENGAH	2,903,145	2,877,448	2,922,516	KALIMANTAN SELATAN 2020	Rp 1	+1
KALIMANTAN SELATAN	2,877,449	2,399,698	2,906,473	KALIMANTAN BARAT 2020	Rp 1	−2
KALIMANTAN TIMUR	2,981,379	2,903,144	3,014,497	KALIMANTAN TENGAH 2020	Rp 1	−2
KALIMANTAN UTARA	3,000,804	2,981,378	3,016,738	KALIMANTAN TIMUR 2020	Rp 1	−1
GORONTALO	2,788,826	4,416,186	2,800,580	— matches nothing in the file

13 of 34 provinces break the bracket. 9 of them are holding another province's wage, to within two rupiah.

the obvious explanation, and why it is wrong A single column that slipped one row is the usual cause of this, and it is the first thing to test. Shifting the whole 2021 column back by one row and re-checking against 2020 makes the agreement worse, not better: +0.2101 against +0.6812 for the file as it stands. The row gaps in the table above are irregular. This is a scramble, not a shift, and the repair is not available from inside the file.

There is a fingerprint. Several displaced values sit exactly one rupiah from their source, and two provinces that otherwise froze their wage are off by exactly one rupiah too (LAMPUNG, SULAWESI TENGGARA). A one-rupiah error is not a typing mistake or a policy decision. It is what a number looks like after a round trip through a floating-point value, which suggests this column reached the file down a different path than the others.

what the year actually was, and what keeping it would have done

provinces with a usable 2021 value	21
of those, froze the wage exactly	18
median people per wage, 2021 as published	5.03
its spread (CV), against 0.223 in a clean year	0.297
its range	2.53 – 11.33

The pandemic freeze is visible in the file only once the corrupted rows are set aside. And the median barely moves, which is why this is dangerous: an analysis that reported only medians would have published 2021 without noticing anything at all.