diff --git a/pandas/core/frame.py b/pandas/core/frame.py
index aeca7782e3ae5..aebf15eb55374 100644
--- a/pandas/core/frame.py
+++ b/pandas/core/frame.py
@@ -5980,7 +5980,7 @@ def pivot_table(
         margins=False,
         dropna=True,
         margins_name="All",
-        observed=False,
+        observed=True,
     ):
         from pandas.core.reshape.pivot import pivot_table
 
diff --git a/pandas/core/reshape/pivot.py b/pandas/core/reshape/pivot.py
index d653dd87308cf..dbfefc1fa7f4a 100644
--- a/pandas/core/reshape/pivot.py
+++ b/pandas/core/reshape/pivot.py
@@ -29,7 +29,7 @@ def pivot_table(
     margins=False,
     dropna=True,
     margins_name="All",
-    observed=False,
+    observed=True,
 ):
     index = _convert_by(index)
     columns = _convert_by(columns)
@@ -92,11 +92,12 @@ def pivot_table(
                 pass
         values = list(values)
 
-    grouped = data.groupby(keys, observed=observed)
+    if dropna:
+        grouped = data.groupby(keys, observed=observed)
+    else:
+        grouped = data.groupby(keys, observed=False)
     agged = grouped.agg(aggfunc)
     if dropna and isinstance(agged, ABCDataFrame) and len(agged.columns):
-        agged = agged.dropna(how="all")
-
         # gh-21133
         # we want to down cast if
         # the original values are ints
@@ -172,10 +173,6 @@ def pivot_table(
     if len(index) == 0 and len(columns) > 0:
         table = table.T
 
-    # GH 15193 Make sure empty columns are removed if dropna=True
-    if isinstance(table, ABCDataFrame) and dropna:
-        table = table.dropna(how="all", axis=1)
-
     return table
 
 
diff --git a/pandas/tests/reshape/test_pivot.py b/pandas/tests/reshape/test_pivot.py
index 582084e3bfb5a..ba48bd70bd807 100644
--- a/pandas/tests/reshape/test_pivot.py
+++ b/pandas/tests/reshape/test_pivot.py
@@ -185,6 +185,59 @@ def test_pivot_table_dropna(self):
         tm.assert_index_equal(pv_col.columns, m)
         tm.assert_index_equal(pv_ind.index, m)
 
+    def test_pivot_table_keep_nancols(self):  # GH18030
+        df = pd.DataFrame(
+            {
+                "metric_value": [10, 11, 0, 3, np.nan, np.nan, 100, 20],
+                "metric_name": ["m", "n", "m", "x", "n", "x", "m", "n"],
+                "product": ["A", "A", "B", "B", "C", "C", "D", "D"],
+                "measurer": ["Tom", "Tom", "Bill", "Tom", "Bill", "Tom", "Bill", "Tom"],
+            }
+        )
+        pv_col = df.pivot_table(
+            "metric_value",
+            "metric_name",
+            ["measurer", "product"],
+            dropna=True,
+        )
+        pv_ind = df.pivot_table(
+            "metric_value",
+            ["measurer", "product"],
+            "metric_name",
+            dropna=True,
+        )
+
+        m = MultiIndex.from_tuples(
+            [
+                ("Bill", "B"),
+                ("Bill", "C"),
+                ("Bill", "D"),
+                ("Tom", "A"),
+                ("Tom", "B"),
+                ("Tom", "C"),
+                ("Tom", "D"),
+            ],
+            names=["measurer", "product"],
+        )
+        tm.assert_index_equal(pv_col.columns, m)
+        tm.assert_index_equal(pv_ind.index, m)
+
+        expected_pv_col = pd.DataFrame(
+            {
+                ("Bill", "B"): {"m": 0.0, "n": np.nan, "x": np.nan},
+                ("Bill", "C"): {"m": np.nan, "n": np.nan, "x": np.nan},
+                ("Bill", "D"): {"m": 100.0, "n": np.nan, "x": np.nan},
+                ("Tom", "A"): {"m": 10.0, "n": 11.0, "x": np.nan},
+                ("Tom", "B"): {"m": np.nan, "n": np.nan, "x": 3.0},
+                ("Tom", "C"): {"m": np.nan, "n": np.nan, "x": np.nan},
+                ("Tom", "D"): {"m": np.nan, "n": 20.0, "x": np.nan},
+            }
+        )
+        expected_pv_col.index.name = "metric_name"
+        expected_pv_col.columns.names = ["measurer", "product"]
+
+        tm.assert_frame_equal(pv_col, expected_pv_col)
+
     def test_pivot_table_categorical(self):
 
         cat1 = Categorical(