diff --git a/doc/source/whatsnew/v0.19.2.txt b/doc/source/whatsnew/v0.19.2.txt
index 978e7616c1018..f193de7fbdbd0 100644
--- a/doc/source/whatsnew/v0.19.2.txt
+++ b/doc/source/whatsnew/v0.19.2.txt
@@ -33,6 +33,7 @@ Bug Fixes
 
 
 - Bug in ``pd.cut`` with negative values and a single bin (:issue:`14652`)
+- Bug in ``pd.to_numeric`` where a 0 was not unsigned on a ``downcast='unsigned'`` argument (:issue:`14401`)
 
 
 
diff --git a/pandas/tools/tests/test_util.py b/pandas/tools/tests/test_util.py
index 8c16308d79a31..f9647721e3c5b 100644
--- a/pandas/tools/tests/test_util.py
+++ b/pandas/tools/tests/test_util.py
@@ -4,9 +4,10 @@
 import nose
 
 import numpy as np
+from numpy import iinfo
 
 import pandas as pd
-from pandas import date_range, Index
+from pandas import (date_range, Index, _np_version_under1p9)
 import pandas.util.testing as tm
 from pandas.tools.util import cartesian_product, to_numeric
 
@@ -401,6 +402,41 @@ def test_downcast(self):
             res = pd.to_numeric(data, downcast=downcast)
             tm.assert_numpy_array_equal(res, expected)
 
+    def test_downcast_limits(self):
+        # Test the limits of each downcast. Bug: #14401.
+        # Check to make sure numpy is new enough to run this test.
+        if _np_version_under1p9:
+            raise nose.SkipTest("Numpy version is under 1.9")
+
+        i = 'integer'
+        u = 'unsigned'
+        dtype_downcast_min_max = [
+            ('int8', i, [iinfo(np.int8).min, iinfo(np.int8).max]),
+            ('int16', i, [iinfo(np.int16).min, iinfo(np.int16).max]),
+            ('int32', i, [iinfo(np.int32).min, iinfo(np.int32).max]),
+            ('int64', i, [iinfo(np.int64).min, iinfo(np.int64).max]),
+            ('uint8', u, [iinfo(np.uint8).min, iinfo(np.uint8).max]),
+            ('uint16', u, [iinfo(np.uint16).min, iinfo(np.uint16).max]),
+            ('uint32', u, [iinfo(np.uint32).min, iinfo(np.uint32).max]),
+            # Test will be skipped until there is more uint64 support.
+            # ('uint64', u, [iinfo(uint64).min, iinfo(uint64).max]),
+            ('int16', i, [iinfo(np.int8).min, iinfo(np.int8).max + 1]),
+            ('int32', i, [iinfo(np.int16).min, iinfo(np.int16).max + 1]),
+            ('int64', i, [iinfo(np.int32).min, iinfo(np.int32).max + 1]),
+            ('int16', i, [iinfo(np.int8).min - 1, iinfo(np.int16).max]),
+            ('int32', i, [iinfo(np.int16).min - 1, iinfo(np.int32).max]),
+            ('int64', i, [iinfo(np.int32).min - 1, iinfo(np.int64).max]),
+            ('uint16', u, [iinfo(np.uint8).min, iinfo(np.uint8).max + 1]),
+            ('uint32', u, [iinfo(np.uint16).min, iinfo(np.uint16).max + 1]),
+            # Test will be skipped until there is more uint64 support.
+            # ('uint64', u, [iinfo(np.uint32).min, iinfo(np.uint32).max + 1]),
+        ]
+
+        for dtype, downcast, min_max in dtype_downcast_min_max:
+            series = pd.to_numeric(pd.Series(min_max), downcast=downcast)
+            tm.assert_equal(series.dtype, dtype)
+
+
 if __name__ == '__main__':
     nose.runmodule(argv=[__file__, '-vvs', '-x', '--pdb', '--pdb-failure'],
                    exit=False)
diff --git a/pandas/tools/util.py b/pandas/tools/util.py
index fec56328c1721..b50bf9dc448bc 100644
--- a/pandas/tools/util.py
+++ b/pandas/tools/util.py
@@ -205,7 +205,7 @@ def to_numeric(arg, errors='raise', downcast=None):
 
         if downcast in ('integer', 'signed'):
             typecodes = np.typecodes['Integer']
-        elif downcast == 'unsigned' and np.min(values) > 0:
+        elif downcast == 'unsigned' and np.min(values) >= 0:
             typecodes = np.typecodes['UnsignedInteger']
         elif downcast == 'float':
             typecodes = np.typecodes['Float']