diff --git a/changes/364.bugfix.rst b/changes/364.bugfix.rst
new file mode 100644
index 00000000..96801fac
--- /dev/null
+++ b/changes/364.bugfix.rst
@@ -0,0 +1 @@
+Allow ``merge_property_trees`` to retain input schema id in ``model.schema["id"]``.
diff --git a/src/stdatamodels/jwst/datamodels/tests/test_models.py b/src/stdatamodels/jwst/datamodels/tests/test_models.py
index cedd4ba2..5a76a00b 100644
--- a/src/stdatamodels/jwst/datamodels/tests/test_models.py
+++ b/src/stdatamodels/jwst/datamodels/tests/test_models.py
@@ -129,7 +129,7 @@ def test_imagemodel():
 
 
 def test_model_with_nonstandard_primary_array():
-    class NonstandardPrimaryArrayModel(JwstDataModel):
+    class _NonstandardPrimaryArrayModel(JwstDataModel):
         schema_url = os.path.join(ROOT_DIR, "nonstandard_primary_array.schema.yaml")
 
         # The wavelength array is the primary array.
@@ -137,7 +137,7 @@ class NonstandardPrimaryArrayModel(JwstDataModel):
         def get_primary_array_name(self):
             return 'wavelength'
 
-    m = NonstandardPrimaryArrayModel((10,))
+    m = _NonstandardPrimaryArrayModel((10,))
     assert 'wavelength' in list(m.keys())
     assert m.wavelength.sum() == 0
 
diff --git a/src/stdatamodels/schema.py b/src/stdatamodels/schema.py
index 1edebdd5..70c59a0f 100644
--- a/src/stdatamodels/schema.py
+++ b/src/stdatamodels/schema.py
@@ -1,7 +1,6 @@
 # Licensed under a 3-clause BSD style license - see LICENSE.rst
 
 import re
-from collections import OrderedDict
 
 
 # return_result included for backward compatibility
@@ -169,11 +168,17 @@ def merge_property_trees(schema):
     can be represented in individual files and then referenced elsewhere. They
     are then combined by this function into a single schema data structure.
     """
-    newschema = OrderedDict()
+    # track the "combined" and "top" items separately
+    # this allows the top level "id", "$schema", etc to overwrite
+    # combined items (so that the schema["id"] doesn't change).
+    combined_items = {}
+    top_items = {}
 
     def add_entry(path, schema, combiner):
+        if not top_items:
+            top_items.update(schema)
         # TODO: Simplify?
-        cursor = newschema
+        cursor = combined_items
         for i in range(len(path)):
             part = path[i]
             if part == combiner:
@@ -185,23 +190,23 @@ def add_entry(path, schema, combiner):
                     cursor.append({})
                 cursor = cursor[part]
             elif part == 'items':
-                cursor = cursor.setdefault('items', OrderedDict())
+                cursor = cursor.setdefault('items', {})
             else:
-                cursor = cursor.setdefault('properties', OrderedDict())
+                cursor = cursor.setdefault('properties', {})
                 if i < len(path) - 1 and isinstance(path[i + 1], int):
                     cursor = cursor.setdefault(part, [])
                 else:
-                    cursor = cursor.setdefault(part, OrderedDict())
+                    cursor = cursor.setdefault(part, {})
 
         cursor.update(schema)
 
     def callback(schema, path, combiner, ctx, recurse):
-        type = schema.get('type')
-        schema = OrderedDict(schema)
-        if type == 'object':
+        schema_type = schema.get('type')
+        schema = dict(schema)  # shallow copy
+        if schema_type == 'object':
             if 'properties' in schema:
                 del schema['properties']
-        elif type == 'array':
+        elif schema_type == 'array':
             del schema['items']
         if 'allOf' in schema:
             del schema['allOf']
@@ -210,7 +215,7 @@ def callback(schema, path, combiner, ctx, recurse):
 
     walk_schema(schema, callback)
 
-    return newschema
+    return combined_items | top_items
 
 
 def build_docstring(klass, template="{fits_hdu} {title}"):
diff --git a/tests/test_schema.py b/tests/test_schema.py
index 519834ca..b3cdbade 100644
--- a/tests/test_schema.py
+++ b/tests/test_schema.py
@@ -206,6 +206,19 @@ def test_merge_property_trees(combiner):
     assert f == s
 
 
+def test_merge_property_tree_top():
+    s = {
+        "id": "foo",
+        "allOf": [
+            {
+                "id": "bar",
+            },
+        ]
+    }
+    f = merge_property_trees(s)
+    assert f["id"] == "foo"
+
+
 def test_schema_docstring():
     template = "{fits_hdu} {title}"
     docstring = build_docstring(FitsModel, template).split("\n")