diff --git a/doc/source/whatsnew/v0.18.1.txt b/doc/source/whatsnew/v0.18.1.txt
index 70a1ad4a335ea..24bd6101e96aa 100644
--- a/doc/source/whatsnew/v0.18.1.txt
+++ b/doc/source/whatsnew/v0.18.1.txt
@@ -43,3 +43,4 @@ Performance Improvements
 
 Bug Fixes
 ~~~~~~~~~
+- Bug in ``value_counts`` where normalizes over all observations including missing even when ``dropna=True`` (:issue:`12558`)
diff --git a/pandas/core/algorithms.py b/pandas/core/algorithms.py
index ad1efa21e8280..6661017ba1b4a 100644
--- a/pandas/core/algorithms.py
+++ b/pandas/core/algorithms.py
@@ -342,7 +342,14 @@ def value_counts(values, sort=True, ascending=False, normalize=False,
         result = result.sort_values(ascending=ascending)
 
     if normalize:
-        result = result / float(values.size)
+        if dropna:
+            # NaT's dropped above if time, so don't need to do again.
+            if not com.is_datetime_or_timedelta_dtype(dtype) \
+                    and not is_period and not is_datetimetz:
+
+                result = result / float(Series(values).count())
+        else:
+            result = result / float(values.size)
 
     return result
 
diff --git a/pandas/tests/test_algos.py b/pandas/tests/test_algos.py
index b9a4384603cda..20d8a3b088611 100644
--- a/pandas/tests/test_algos.py
+++ b/pandas/tests/test_algos.py
@@ -517,6 +517,33 @@ def test_dropna(self):
                 pd.Series([10.3, 5., 5., None]).value_counts(dropna=False),
                 pd.Series([2, 1, 1], index=[5., 10.3, np.nan]))
 
+    def test_normalize(self):
+        # Issue 12558
+        tm.assert_series_equal(
+            pd.Series([5., 10.3, 10.3, 10.3, np.nan]).value_counts(
+                dropna=True, normalize=True),
+            pd.Series([0.75, 0.25], index=[10.3, 5.]))
+
+        tm.assert_series_equal(
+            pd.Series([5., 10.3, 10.3, 10.3, np.nan]).value_counts(
+                dropna=True, normalize=False),
+            pd.Series([3, 1], index=[10.3, 5.]))
+
+        tm.assert_series_equal(
+            pd.Series([5., 10.3, 10.3, 10.3, np.nan]).value_counts(
+                dropna=False, normalize=True),
+            pd.Series([0.6, 0.2, 0.2], index=[10.3, 5., np.nan]))
+
+        tm.assert_series_equal(
+            pd.Series([5., 10.3, 10.3, 10.3, np.nan]).value_counts(
+                dropna=False, normalize=False),
+            pd.Series([3, 1, 1], index=[10.3, 5., np.nan]))
+
+        s = pd.Series(['2015-01-03T00:00:00.000000000+0000', pd.NaT])
+        tm.assert_series_equal(
+            s.value_counts(dropna=True, normalize=True),
+            pd.Series([1], index=[s.loc[0]]))
+
 
 class GroupVarTestMixin(object):