BUG: value_counts returning incorrect dtype for string dtype

doc/source/whatsnew/v2.1.2.rst

@@ -23,6 +23,7 @@ Fixed regressions
 
 Bug fixes
 ~~~~~~~~~
+- Fixed bug in :meth:`.SeriesGroupBy.value_counts` returning incorrect dtype for string columns (:issue:`55626`)
 - Fixed bug in :meth:`Categorical.equals` if other has arrow backed string dtype (:issue:`55364`)
 - Fixed bug in :meth:`DataFrame.__setitem__` not inferring string dtype for zero-dimensional array with ``infer_string=True`` (:issue:`55366`)
 - Fixed bug in :meth:`DataFrame.idxmin` and :meth:`DataFrame.idxmax` raising for arrow dtypes (:issue:`55368`)

pandas/core/groupby/groupby.py

@@ -108,6 +108,10 @@ class providing the base-class of operations.
     SparseArray,
 )
 from pandas.core.arrays.string_ import StringDtype
+from pandas.core.arrays.string_arrow import (
+    ArrowStringArray,
+    ArrowStringArrayNumpySemantics,
+)
 from pandas.core.base import (
     PandasObject,
     SelectionMixin,
@@ -2855,7 +2859,9 @@ def _value_counts(
             result_series.name = name
             result_series.index = index.set_names(range(len(columns)))
             result_frame = result_series.reset_index()
-            result_frame.columns = columns + [name]
+            result_frame.columns = Index(
+                columns + [name], dtype=self.grouper.groupings[0].obj.columns.dtype
+            )
             result = result_frame
         return result.__finalize__(self.obj, method="value_counts")
 
@@ -3007,7 +3013,12 @@ def size(self) -> DataFrame | Series:
         dtype_backend: None | Literal["pyarrow", "numpy_nullable"] = None
         if isinstance(self.obj, Series):
             if isinstance(self.obj.array, ArrowExtensionArray):
-                dtype_backend = "pyarrow"
+                if isinstance(self.obj.array, ArrowStringArrayNumpySemantics):
+                    dtype_backend = None
+                elif isinstance(self.obj.array, ArrowStringArray):
+                    dtype_backend = "numpy_nullable"
+                else:
+                    dtype_backend = "pyarrow"
             elif isinstance(self.obj.array, BaseMaskedArray):
                 dtype_backend = "numpy_nullable"
         # TODO: For DataFrames what if columns are mixed arrow/numpy/masked?

pandas/tests/groupby/methods/test_value_counts.py

@@ -9,6 +9,8 @@
 import numpy as np
 import pytest
 
+from pandas.compat import pa_version_under7p0
+
 from pandas import (
     Categorical,
     CategoricalIndex,
@@ -369,6 +371,20 @@ def test_against_frame_and_seriesgroupby(
             tm.assert_frame_equal(result, expected)
 
 
+@pytest.mark.parametrize(
+    "dtype",
+    [
+        object,
+        pytest.param(
+            "string[pyarrow_numpy]",
+            marks=pytest.mark.skipif(pa_version_under7p0, reason="arrow not installed"),
+        ),
+        pytest.param(
+            "string[pyarrow]",
+            marks=pytest.mark.skipif(pa_version_under7p0, reason="arrow not installed"),
+        ),
+    ],
+)
 @pytest.mark.parametrize("normalize", [True, False])
 @pytest.mark.parametrize(
     "sort, ascending, expected_rows, expected_count, expected_group_size",
@@ -386,7 +402,10 @@ def test_compound(
     expected_rows,
     expected_count,
     expected_group_size,
+    dtype,
 ):
+    education_df = education_df.astype(dtype)
+    education_df.columns = education_df.columns.astype(dtype)
     # Multiple groupby keys and as_index=False
     gp = education_df.groupby(["country", "gender"], as_index=False, sort=False)
     result = gp["education"].value_counts(
@@ -395,11 +414,17 @@ def test_compound(
     expected = DataFrame()
     for column in ["country", "gender", "education"]:
         expected[column] = [education_df[column][row] for row in expected_rows]
+        expected = expected.astype(dtype)
+        expected.columns = expected.columns.astype(dtype)
     if normalize:
         expected["proportion"] = expected_count
         expected["proportion"] /= expected_group_size
+        if dtype == "string[pyarrow]":
+            expected["proportion"] = expected["proportion"].convert_dtypes()
     else:
         expected["count"] = expected_count
+        if dtype == "string[pyarrow]":
+            expected["count"] = expected["count"].convert_dtypes()
     tm.assert_frame_equal(result, expected)