REGR: .describe on unsigned dtypes results in object

pandas/core/describe.py

@@ -240,7 +240,12 @@ def describe_numeric_1d(series: Series, percentiles: Sequence[float]) -> Series:
         + series.quantile(percentiles).tolist()
         + [series.max()]
     )
-    return Series(d, index=stat_index, name=series.name)
+
+    result = Series(d, index=stat_index, name=series.name)
+    if is_numeric_dtype(series):
+        # GH#48340 - don't rely on inference, always return float on numeric data
+        result = result.astype(float)
+    return result
 
 
 def describe_categorical_1d(

pandas/tests/series/methods/test_describe.py

@@ -1,4 +1,5 @@
 import numpy as np
+import pytest
 
 from pandas import (
     Period,
@@ -149,3 +150,25 @@ def test_datetime_is_numeric_includes_datetime(self):
             index=["count", "mean", "min", "25%", "50%", "75%", "max"],
         )
         tm.assert_series_equal(result, expected)
+
+    @pytest.mark.parametrize(
+        "dtype", ["int32", "int64", "uint32", "uint64", "float32", "float64"]
+    )
+    def test_numeric_result_is_float(self, dtype):
+        # GH#48340 - describe should always return dtype float on numeric input
+        ser = Series([0, 1], dtype=dtype)
+        result = ser.describe()
+        expected = Series(
+            [
+                2.0,
+                0.5,
+                ser.std(),
+                0,
+                0.25,
+                0.5,
+                0.75,
+                1.0,
+            ],
+            index=["count", "mean", "std", "min", "25%", "50%", "75%", "max"],
+        )
+        tm.assert_series_equal(result, expected)