diff --git a/python/pyspark/pandas/tests/indexes/test_datetime.py b/python/pyspark/pandas/tests/indexes/test_datetime.py
index 85a2b21901774..053e7f341222a 100644
--- a/python/pyspark/pandas/tests/indexes/test_datetime.py
+++ b/python/pyspark/pandas/tests/indexes/test_datetime.py
@@ -240,6 +240,11 @@ def test_map(self):
         mapper_pser = pd.Series([1, 2, 3], index=pidx)
         self.assert_eq(psidx.map(mapper_pser), pidx.map(mapper_pser))
 
+    def test_repr(self):
+        pidx_repr = pd.DatetimeIndex(['1970-01-01', '1970-02-01', '1970-03-01']).__repr__()
+        psidx_repr = ps.DatetimeIndex(['1970-01-01', '1970-02-01', '1970-03-01']).__repr__()
+        self.assert_eq(pidx_repr, psidx_repr)
+
 
 if __name__ == "__main__":
     import unittest
diff --git a/python/pyspark/sql/pandas/conversion.py b/python/pyspark/sql/pandas/conversion.py
index 119a9bf315caa..ce3c7886bd2fc 100644
--- a/python/pyspark/sql/pandas/conversion.py
+++ b/python/pyspark/sql/pandas/conversion.py
@@ -87,7 +87,7 @@ def toPandas(self) -> "PandasDataFrameLike":
 
         import numpy as np
         import pandas as pd
-        from pandas.core.dtypes.common import is_timedelta64_dtype
+        from pandas.core.dtypes.common import is_timedelta64_dtype, is_datetime64_dtype
 
         jconf = self.sparkSession._jconf
         timezone = jconf.sessionLocalTimeZone()
@@ -244,7 +244,10 @@ def toPandas(self) -> "PandasDataFrameLike":
             # No need to cast for non-empty series for timedelta. The type is already correct.
             should_check_timedelta = is_timedelta64_dtype(t) and len(pdf) == 0
 
-            if (t is not None and not is_timedelta64_dtype(t)) or should_check_timedelta:
+            if (t is not None and
+                    not all([
+                        is_timedelta64_dtype(t),
+                        is_datetime64_dtype(t)])) or should_check_timedelta:
                 series = series.astype(t, copy=False)
 
             with catch_warnings():