Address some PR feedback

Author: wisznie

pandas/core/tools/numeric.py

@@ -222,11 +222,8 @@ def to_numeric(
 
     values_dtype = getattr(values, "dtype", None)
     if isinstance(values_dtype, ArrowDtype):
-        if is_numeric_dtype(values_dtype):
-            if is_series:
-                return arg._constructor(values, index=arg.index, name=arg.name)
-            else:
-                return values
+        if is_numeric_dtype(values_dtype) and is_series:
+            return arg._constructor(values, index=arg.index, name=arg.name)
 
         mask = values.isna()
         values = values.dropna().to_numpy()

pandas/tests/tools/test_to_numeric.py

@@ -921,12 +921,18 @@ def test_coerce_pyarrow_backend():
     tm.assert_series_equal(result, expected)
 
 
-def test_to_numeric_arrow_decimal_with_na():
+@pytest.mark.parametrize(
+    "dtype",
+    [
+        "ArrowDtype",
+    ],
+)
+def test_to_numeric_arrow_decimal_with_na(dtype):
     # GH 61641
     pa = pytest.importorskip("pyarrow")
-    decimal_type = ArrowDtype(pa.decimal128(3, scale=2))
+    target_class = globals()[dtype]
+    decimal_type = target_class(pa.decimal128(3, scale=2))
     series = Series([1, None], dtype=decimal_type)
     result = to_numeric(series, errors="coerce")
 
-    expected = Series([1.00, pd.NA], dtype=decimal_type)
-    tm.assert_series_equal(result, expected)
+    tm.assert_series_equal(result, series)