diff --git a/tests/unittests/test_scipy.py b/tests/unittests/test_scipy.py
index 185aaa0..18c9182 100644
--- a/tests/unittests/test_scipy.py
+++ b/tests/unittests/test_scipy.py
@@ -279,7 +279,7 @@ def test_recall_diff_hybrid():
     test_recall_diff('hybrid')
 
 
-# """ too slow and disabled
+""" too slow -- keep disabled unless hunting non-deterministic bugs.
 def test_many_recall_diff_hybrid(many=2000):
     count = 0
     for i in range(many):
@@ -421,3 +421,43 @@ def test_frontier_slow():
 
 def test_frontier_hybrid():
     test_frontier('hybrid')
+
+
+def test_selection_rate_diff_levelling_up(use_fast=True):
+    """Maximize accuracy while enforcing demographic parity with levelling up."""
+
+    fpredictor = fair.FairPredictor(
+        predictor, test_dict, "sex_ Female", use_fast=use_fast)
+
+    fpredictor.fit(gm.accuracy, gm.pos_pred_rate.diff, 0.025, force_levelling_up=True)
+
+    rate = fpredictor.evaluate_groups(metrics={1: gm.pos_pred_rate}, verbose=False)[1]
+    assert (rate['updated'].drop('Maximum difference') >= rate['original'].drop('Maximum difference')).all()
+    assert (fpredictor.evaluate_groups() == fpredictor.evaluate_groups(test_dict)).all().all()
+    # Evaluate the change in fairness (recall difference corresponds to EO)
+    measures = fpredictor.evaluate_fairness(verbose=False)
+
+    assert measures["original"]["statistical_parity"] > 0.025
+
+    assert measures["updated"]["statistical_parity"] <= 0.025
+
+    fpredictor.fit(gm.accuracy, gm.pos_pred_rate.diff, 0.025, force_levelling_up='-')
+    rate = fpredictor.evaluate_groups(metrics={1: gm.pos_pred_rate}, verbose=False)[1]
+    assert (rate['updated'].drop('Maximum difference') <= rate['original'].drop('Maximum difference')).all()
+    # Evaluate the change in fairness (recall difference corresponds to EO)
+    measures = fpredictor.evaluate_fairness(verbose=False)
+
+    # Evaluate the change in fairness (recall difference corresponds to EO)
+    measures = fpredictor.evaluate_fairness(verbose=False)
+
+    assert measures["original"]["statistical_parity"] > 0.025
+
+    assert measures["updated"]["statistical_parity"] <= 0.025
+
+
+def test_selection_rate_diff_levelling_up_slow():
+    test_selection_rate_diff_levelling_up(use_fast=False)
+
+
+def test_selection_rate_diff_levelling_up_hybrid():
+    test_selection_rate_diff_levelling_up(use_fast='hybrid')
\ No newline at end of file