BUG: do not cast ints to floats if inputs o crosstab are not aligned

closes pandas-dev#17005

doc/source/whatsnew/v0.21.0.txt

@@ -221,7 +221,7 @@ Sparse
 Reshaping
 ^^^^^^^^^
 - Joining/Merging with a non unique ``PeriodIndex`` raised a TypeError (:issue:`16871`)
-
+- Bug in :func:`crosstab` where non-aligned series of integers were casted to float (:issue:`17005`)
 
 Numeric
 ^^^^^^^

pandas/core/reshape/pivot.py

@@ -6,6 +6,7 @@
 from pandas import Series, DataFrame, MultiIndex, Index
 from pandas.core.groupby import Grouper
 from pandas.core.reshape.util import cartesian_product
+from pandas.core.index import _get_combined_index
 from pandas.compat import range, lrange, zip
 from pandas import compat
 import pandas.core.common as com
@@ -493,6 +494,13 @@ def crosstab(index, columns, values=None, rownames=None, colnames=None,
     rownames = _get_names(index, rownames, prefix='row')
     colnames = _get_names(columns, colnames, prefix='col')
 
+    obs_idxes = [obj.index for objs in (index, columns) for obj in objs
+                 if hasattr(obj, 'index')]
+    if obs_idxes:
+        common_idx = _get_combined_index(obs_idxes, intersect=True)
+    else:
+        common_idx = None
+
     data = {}
     data.update(zip(rownames, index))
     data.update(zip(colnames, columns))
@@ -503,20 +511,21 @@ def crosstab(index, columns, values=None, rownames=None, colnames=None,
     if values is not None and aggfunc is None:
         raise ValueError("values cannot be used without an aggfunc.")
 
+    df = DataFrame(data, index=common_idx)
     if values is None:
-        df = DataFrame(data)
         df['__dummy__'] = 0
-        table = df.pivot_table('__dummy__', index=rownames, columns=colnames,
-                               aggfunc=len, margins=margins,
-                               margins_name=margins_name, dropna=dropna)
-        table = table.fillna(0).astype(np.int64)
-
+        kwargs = {'aggfunc': len, 'fill_value': 0}
     else:
-        data['__dummy__'] = values
-        df = DataFrame(data)
-        table = df.pivot_table('__dummy__', index=rownames, columns=colnames,
-                               aggfunc=aggfunc, margins=margins,
-                               margins_name=margins_name, dropna=dropna)
+        df['__dummy__'] = values
+        kwargs = {'aggfunc': aggfunc}
+
+    table = df.pivot_table('__dummy__', index=rownames, columns=colnames,
+                           margins=margins, margins_name=margins_name,
+                           dropna=dropna, **kwargs)
+
+    # GH 17013:
+    if values is None and margins:
+        table = table.fillna(0).astype(np.int64)
 
     # Post-process
     if normalize is not False:

pandas/tests/reshape/test_pivot.py

@@ -1058,6 +1058,22 @@ def test_crosstab_ndarray(self):
         assert result.index.name == 'row_0'
         assert result.columns.name == 'col_0'
 
+    def test_crosstab_non_aligned(self):
+        # GH 17005
+        a = pd.Series([0, 1, 1], index=['a', 'b', 'c'])
+        b = pd.Series([3, 4, 3, 4, 3], index=['a', 'b', 'c', 'd', 'f'])
+        c = np.array([3, 4, 3])
+
+        expected = pd.DataFrame([[1, 0], [1, 1]],
+                                index=Index([0, 1], name='row_0'),
+                                columns=Index([3, 4], name='col_0'))
+
+        result = crosstab(a, b)
+        tm.assert_frame_equal(result, expected)
+
+        result = crosstab(a, c)
+        tm.assert_frame_equal(result, expected)
+
     def test_crosstab_margins(self):
         a = np.random.randint(0, 7, size=100)
         b = np.random.randint(0, 3, size=100)