diff --git a/quantecon/timings/tests/__init__.py b/quantecon/timings/tests/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/quantecon/timings/tests/test_timings.py b/quantecon/timings/tests/test_timings.py
new file mode 100644
index 00000000..59622c4d
--- /dev/null
+++ b/quantecon/timings/tests/test_timings.py
@@ -0,0 +1,64 @@
+"""
+Tests for quantecon/timings/timings.py
+
+Covers the previously zero-coverage public API:
+
+* ``float_precision`` getter (precision=None) and setter
+* ``ValueError`` paths for a negative int, a float, and a string
+* ``get_default_precision`` returning the current global value
+* global-state restoration so tests never leak precision changes
+"""
+import pytest
+
+import quantecon as qe
+
+
+@pytest.fixture(autouse=True)
+def restore_precision():
+    """Snapshot the global precision before each test and restore it after."""
+    original = qe.timings.get_default_precision()
+    yield original
+    qe.timings.float_precision(original)
+
+
+def test_get_default_precision_returns_current(restore_precision):
+    # get_default_precision reports the live global value, not a constant.
+    assert qe.timings.get_default_precision() == restore_precision
+
+
+def test_float_precision_getter_returns_current(restore_precision):
+    # Calling with no argument returns the current precision.
+    assert qe.timings.float_precision() == restore_precision
+
+
+def test_float_precision_setter_changes_value():
+    # Setting must be observable via both the getter and get_default_precision.
+    qe.timings.float_precision(6)
+    assert qe.timings.float_precision() == 6
+    assert qe.timings.get_default_precision() == 6
+
+
+def test_float_precision_setter_accepts_zero():
+    # Zero is a valid non-negative integer.
+    qe.timings.float_precision(0)
+    assert qe.timings.float_precision() == 0
+
+
+def test_float_precision_setter_accepts_large_value():
+    qe.timings.float_precision(12)
+    assert qe.timings.float_precision() == 12
+
+
+@pytest.mark.parametrize("bad", [-1, 4.5, "4"])
+def test_float_precision_value_error_on_invalid(bad):
+    # Negative int, float, and string must all raise ValueError.
+    with pytest.raises(ValueError, match="non-negative integer"):
+        qe.timings.float_precision(bad)
+
+
+def test_float_precision_value_error_does_not_change_state():
+    # A failed set must leave the global precision untouched.
+    before = qe.timings.get_default_precision()
+    with pytest.raises(ValueError):
+        qe.timings.float_precision(-1)
+    assert qe.timings.get_default_precision() == before