Pandas version checks
-
[x] I have checked that this issue has not already been reported.
-
[x] I have confirmed this bug exists on the latest version of pandas.
-
[ ] I have confirmed this bug exists on the main branch of pandas.
Reproducible Example
import numpy as np
import pandas as pd
tz = "Australia/Sydney"
start = pd.Timestamp.now(tz).normalize()
index = pd.date_range(start=start, periods=24, freq="h", tz=tz)
rng = np.random.default_rng(42)
values = rng.random(len(index))
df = pd.DataFrame({"value": values}, index=index)
df.index.name = "value_timestamp"
df = df.reset_index()
pa_df = df.convert_dtypes(dtype_backend='pyarrow')
expected = df['value_timestamp'].dt.tz_localize(None)
actual = pa_df['value_timestamp'].dt.tz_localize(None)
assert actual.equals(expected)
Issue Description
tz_localize(None) should strip the timezone information and otherwise leave the timestamps unchanged. This is what happens with the normal dtype backend. With pyarrow dtypes the timestamps are incorrectly shifted by the timestamp offset.
Expected Behavior
The pyarrow dtypes don't shift the timestamps and give the same result as the normal dtypes.
Installed Versions
INSTALLED VERSIONS
------------------
commit : 0691c5cf90477d3503834d983f69350f250a6ff7
python : 3.12.11
python-bits : 64
OS : Linux
OS-release : 6.14.0-1010-aws
Version : #10~24.04.1-Ubuntu SMP Fri Jul 18 20:44:30 UTC 2025
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : C.UTF-8
LOCALE : C.UTF-8
pandas : 2.2.3
numpy : 1.26.4
pytz : 2025.2
dateutil : 2.9.0.post0
pip : 25.2
Cython : None
sphinx : 8.2.3
IPython : 9.4.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.13.4
blosc : None
bottleneck : None
dataframe-api-compat : None
fastparquet : None
fsspec : 2025.7.0
html5lib : 1.1
hypothesis : 6.137.3
gcsfs : None
jinja2 : 3.1.6
lxml.etree : 5.4.0
matplotlib : 3.10.5
numba : None
numexpr : None
odfpy : None
openpyxl : 3.1.5
pandas_gbq : None
psycopg2 : None
pymysql : None
pyarrow : 19.0.1
pyreadstat : None
pytest : 8.4.1
python-calamine : None
pyxlsb : None
s3fs : 2025.7.0
scipy : 1.16.1
sqlalchemy : 2.0.37
tables : None
tabulate : 0.9.0
xarray : 2025.7.1
xlrd : None
xlsxwriter : None
zstandard : 0.23.0
tzdata : 2025.2
qtpy : None
pyqt5 : None
Comment From: dhirschfeld
>>> pd.concat([df['value_timestamp'], expected, actual], keys=['original', 'numpy', 'pyarrow'], axis=1)
original numpy pyarrow
0 2025-08-13 00:00:00+10:00 2025-08-13 00:00:00 2025-08-12 14:00:00
1 2025-08-13 01:00:00+10:00 2025-08-13 01:00:00 2025-08-12 15:00:00
2 2025-08-13 02:00:00+10:00 2025-08-13 02:00:00 2025-08-12 16:00:00
3 2025-08-13 03:00:00+10:00 2025-08-13 03:00:00 2025-08-12 17:00:00
4 2025-08-13 04:00:00+10:00 2025-08-13 04:00:00 2025-08-12 18:00:00
5 2025-08-13 05:00:00+10:00 2025-08-13 05:00:00 2025-08-12 19:00:00
6 2025-08-13 06:00:00+10:00 2025-08-13 06:00:00 2025-08-12 20:00:00
7 2025-08-13 07:00:00+10:00 2025-08-13 07:00:00 2025-08-12 21:00:00
8 2025-08-13 08:00:00+10:00 2025-08-13 08:00:00 2025-08-12 22:00:00
9 2025-08-13 09:00:00+10:00 2025-08-13 09:00:00 2025-08-12 23:00:00
10 2025-08-13 10:00:00+10:00 2025-08-13 10:00:00 2025-08-13 00:00:00
11 2025-08-13 11:00:00+10:00 2025-08-13 11:00:00 2025-08-13 01:00:00
12 2025-08-13 12:00:00+10:00 2025-08-13 12:00:00 2025-08-13 02:00:00
13 2025-08-13 13:00:00+10:00 2025-08-13 13:00:00 2025-08-13 03:00:00
14 2025-08-13 14:00:00+10:00 2025-08-13 14:00:00 2025-08-13 04:00:00
15 2025-08-13 15:00:00+10:00 2025-08-13 15:00:00 2025-08-13 05:00:00
16 2025-08-13 16:00:00+10:00 2025-08-13 16:00:00 2025-08-13 06:00:00
17 2025-08-13 17:00:00+10:00 2025-08-13 17:00:00 2025-08-13 07:00:00
18 2025-08-13 18:00:00+10:00 2025-08-13 18:00:00 2025-08-13 08:00:00
19 2025-08-13 19:00:00+10:00 2025-08-13 19:00:00 2025-08-13 09:00:00
20 2025-08-13 20:00:00+10:00 2025-08-13 20:00:00 2025-08-13 10:00:00
21 2025-08-13 21:00:00+10:00 2025-08-13 21:00:00 2025-08-13 11:00:00
22 2025-08-13 22:00:00+10:00 2025-08-13 22:00:00 2025-08-13 12:00:00
23 2025-08-13 23:00:00+10:00 2025-08-13 23:00:00 2025-08-13 13:00:00