diff --git a/python/pyspark/pandas/tests/data_type_ops/test_udt_ops.py b/python/pyspark/pandas/tests/data_type_ops/test_udt_ops.py
index 60b4153198a34..f4f833ea9cf55 100644
--- a/python/pyspark/pandas/tests/data_type_ops/test_udt_ops.py
+++ b/python/pyspark/pandas/tests/data_type_ops/test_udt_ops.py
@@ -130,6 +130,26 @@ def test_from_to_pandas(self):
         self.assert_eq(pser, psser._to_pandas())
         self.assert_eq(ps.from_pandas(pser), psser)
 
+    def test_with_first_null(self):
+        lst = [None, None, None, SparseVector(1, {0: 0.1})]
+        pser = pd.Series(lst)
+        psser = ps.Series(lst)
+        self.assert_eq(pser, psser._to_pandas())
+        self.assert_eq(ps.from_pandas(pser), psser)
+
+        lst2 = [SparseVector(1, {0: 0.1}), None, None, None]
+        pdf = pd.DataFrame({"a": lst, "b": lst2})
+        psdf = ps.DataFrame({"a": lst, "b": lst2})
+        self.assert_eq(pdf, psdf._to_pandas())
+        self.assert_eq(ps.from_pandas(pdf), psdf)
+
+    def test_with_all_null(self):
+        lst = [None, None, None, None]
+        pser = pd.Series(lst, dtype=object)
+        psser = ps.Series(lst, dtype=object)
+        self.assert_eq(pser, psser._to_pandas())
+        self.assert_eq(ps.from_pandas(pser), psser)
+
     def test_isnull(self):
         self.assert_eq(self.pser.isnull(), self.psser.isnull())
 
diff --git a/python/pyspark/pandas/typedef/typehints.py b/python/pyspark/pandas/typedef/typehints.py
index 4244f5831aa50..48545d124b2d8 100644
--- a/python/pyspark/pandas/typedef/typehints.py
+++ b/python/pyspark/pandas/typedef/typehints.py
@@ -362,8 +362,9 @@ def infer_pd_series_spark_type(
     if dtype == np.dtype("object"):
         if len(pser) == 0 or pser.isnull().all():
             return types.NullType()
-        elif hasattr(pser.iloc[0], "__UDT__"):
-            return pser.iloc[0].__UDT__
+        notnull = pser[pser.notnull()]
+        if hasattr(notnull.iloc[0], "__UDT__"):
+            return notnull.iloc[0].__UDT__
         else:
             return from_arrow_type(pa.Array.from_pandas(pser).type, prefer_timestamp_ntz)
     elif isinstance(dtype, CategoricalDtype):