diff --git a/ci/scripts/install_pandas.sh b/ci/scripts/install_pandas.sh
index 03009b3824f..b8a63632e28 100755
--- a/ci/scripts/install_pandas.sh
+++ b/ci/scripts/install_pandas.sh
@@ -40,7 +40,7 @@ if [ "${pandas}" = "upstream_devel" ]; then
 elif [ "${pandas}" = "nightly" ]; then
   pip install --extra-index-url https://pypi.anaconda.org/scientific-python-nightly-wheels/simple --pre pandas
 elif [ "${pandas}" = "latest" ]; then
-  pip install pandas
+  pip install --upgrade pandas
 else
   pip install pandas=="${pandas}"
 fi
diff --git a/python/pyarrow/tests/parquet/test_basic.py b/python/pyarrow/tests/parquet/test_basic.py
index 94868741f39..e50b2715e3e 100644
--- a/python/pyarrow/tests/parquet/test_basic.py
+++ b/python/pyarrow/tests/parquet/test_basic.py
@@ -728,6 +728,7 @@ def test_parquet_file_too_small(tempdir):
 @pytest.mark.fastparquet
 @pytest.mark.filterwarnings("ignore:RangeIndex:FutureWarning")
 @pytest.mark.filterwarnings("ignore:tostring:DeprecationWarning:fastparquet")
+@pytest.mark.filterwarnings("ignore:unclosed file:ResourceWarning")
 def test_fastparquet_cross_compatibility(tempdir):
     fp = pytest.importorskip('fastparquet')
 
@@ -751,17 +752,21 @@ def test_fastparquet_cross_compatibility(tempdir):
 
     fp_file = fp.ParquetFile(file_arrow)
     df_fp = fp_file.to_pandas()
-    tm.assert_frame_equal(df, df_fp)
+    # check_dtype=False: string/categorical dtype handling differs between libraries
+    tm.assert_frame_equal(df, df_fp, check_dtype=False, check_categorical=False)
 
     # Fastparquet -> arrow
     file_fastparquet = str(tempdir / "cross_compat_fastparquet.parquet")
-    fp.write(file_fastparquet, df)
+    # fastparquet can't write pandas 3.0 StringDtype
+    df_for_fp = df.copy()
+    df_for_fp['a'] = df_for_fp['a'].astype(object)
+    df_for_fp['f'] = df_for_fp['f'].astype(object)
+    fp.write(file_fastparquet, df_for_fp)
 
     table_fp = pq.read_pandas(file_fastparquet)
     # for fastparquet written file, categoricals comes back as strings
     # (no arrow schema in parquet metadata)
-    df['f'] = df['f'].astype(object)
-    tm.assert_frame_equal(table_fp.to_pandas(), df)
+    tm.assert_frame_equal(table_fp.to_pandas(), df_for_fp, check_dtype=False)
 
 
 @pytest.mark.parametrize('array_factory', [
diff --git a/python/pyarrow/tests/test_pandas.py b/python/pyarrow/tests/test_pandas.py
index cecf10f2165..e7b625422f4 100644
--- a/python/pyarrow/tests/test_pandas.py
+++ b/python/pyarrow/tests/test_pandas.py
@@ -4956,7 +4956,7 @@ def test_timestamp_as_object_non_nanosecond(resolution, tz, dt):
         assert isinstance(result[0], datetime)
         if tz:
             assert result[0].tzinfo is not None
-            expected = result[0].tzinfo.fromutc(dt)
+            expected = dt.replace(tzinfo=timezone.utc).astimezone(result[0].tzinfo)
         else:
             assert result[0].tzinfo is None
             expected = dt