diff --git a/pandas/tests/indexes/common.py b/pandas/tests/indexes/common.py
index 1b373baf9b3c1..dd4e0869e9627 100644
--- a/pandas/tests/indexes/common.py
+++ b/pandas/tests/indexes/common.py
@@ -127,6 +127,13 @@ def test_ndarray_compat_properties(self):
 
         values = idx.values
         for prop in self._compat_props:
+            # skip equivalency check if converted type is object, as happens
+            # for PeriodIndex, since then object (address) would be 4 bytes
+            # on 32bit platforms and equivalence to int64 of original
+            # date time is just accidental
+            if prop in ('itemsize', 'nbytes') \
+                    and values.dtype.name == 'object':
+                continue
             self.assertEqual(getattr(idx, prop), getattr(values, prop))
 
         # test for validity
diff --git a/pandas/tools/tests/test_tile.py b/pandas/tools/tests/test_tile.py
index 5c7cee862ccd3..4fd53c1f696af 100644
--- a/pandas/tools/tests/test_tile.py
+++ b/pandas/tools/tests/test_tile.py
@@ -288,7 +288,10 @@ def test_qcut_duplicates_bin(self):
 
     def test_single_bin(self):
         # issue 14652
-        expected = Series([0, 0])
+        # Explicit dtype since Series produces int64 for ints, while cut
+        # (due to numpy.searchsorted) would use int32 on i386, so let's assure
+        # correct default to the architecture int
+        expected = Series([0, 0], dtype='intp')
 
         s = Series([9., 9.])
         result = cut(s, 1, labels=False)