diff --git a/CHANGELOG.md b/CHANGELOG.md
index baa9ec8..5dc6d45 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -9,10 +9,20 @@ This project uses [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
 
+## [0.6.5] — 2026-09-24
+
 ### Security
 
 - **The Census API key is no longer written to logs.** `geoinfo_from_params()` logged its request parameters at INFO after adding `key`, so the key appeared in notebook outputs (hundreds of times per call through the hierarchical geography lookup). It now logs before adding the key. Keys passed on to `morpc.req` are redacted there from morpc 0.7.5 (morpc/morpc-py#207).
 
+### Fixed
+
+- **A Census API response of 204 No Content is treated as no rows** instead of failing the whole `CensusAPI` fetch. The Census API returns 204 when a request is valid but it has no data for the requested geographies; for example, ACS 5-year 2017 lists place/remainder parts (`070`) but publishes none. A warning is logged and the result has no rows for that request. Other HTTP errors still raise.
+
+### Changed
+
+- Requires morpc 0.7.5 or later, for its `HTTPError` response and API key redaction.
+
 ## [0.6.4] — 2026-09-23
 
 ### Fixed
diff --git a/morpc_census/api.py b/morpc_census/api.py
index 3a5ba24..373e8f5 100644
--- a/morpc_census/api.py
+++ b/morpc_census/api.py
@@ -934,6 +934,7 @@ def _fetch_variables(self, url: str, params: dict) -> pd.DataFrame:
         and all batch results are joined on GEO_ID into a single DataFrame.
         """
         from morpc.req import get_json_safely
+        from requests import HTTPError
 
         BATCH_SIZE = 48
         variables = self.variables
@@ -959,7 +960,15 @@ def _fetch_variables(self, url: str, params: dict) -> pd.DataFrame:
             batch_params = {**params, 'get': ','.join(['GEO_ID', 'NAME'] + batch)}
             chunk_frames = []
             for geo_chunk in geo_chunks:
-                records = get_json_safely(url, params={**batch_params, **geo_chunk})
+                try:
+                    records = get_json_safely(url, params={**batch_params, **geo_chunk})
+                except HTTPError as e:
+                    # 204 No Content: the request is valid but Census has no data for these geographies
+                    # (e.g. ACS 5-year 2017 lists place/remainder parts but publishes none).
+                    if e.response is None or e.response.status_code != 204:
+                        raise
+                    self.logger.warning(f"Census API returned no data for {self.name}; treating the request as no rows.")
+                    records = [['GEO_ID', 'NAME'] + batch]
                 columns = records.pop(0)
                 chunk_frames.append(pd.DataFrame.from_records(records, columns=columns))
             frames.append(pd.concat(chunk_frames).set_index(['GEO_ID', 'NAME']))
diff --git a/pyproject.toml b/pyproject.toml
index 517dd2d..4903158 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -14,7 +14,7 @@ classifiers = [
     "Typing :: Typed",
 ]
 dependencies = [
-    "morpc>=0.5.4",
+    "morpc>=0.7.5",
     "numpy>=1.24",
     "pandas>=2.0",
     "geopandas>=0.14",
diff --git a/reference/dev_notes.md b/reference/dev_notes.md
index 2fb1357..c34e3a6 100644
--- a/reference/dev_notes.md
+++ b/reference/dev_notes.md
@@ -1106,3 +1106,11 @@ Tests: `tests/test_geos_hierarchical.py` — 3 new tests (scope filtering, one r
 **Note**: `dec/pl` does not publish 070 in any year; `dec/dhc` (2020) and `dec/sf1` (2010/2000) do. Their county subdivision totals match `dec/pl` except where parts of places that no longer exist are missing from the current-geography lookup (e.g. Hidden Lakes CDP).
 
 Tests: 1 new in `tests/test_geos_hierarchical.py`, 2 new in `tests/test_api.py`. 343 passing. Live: region15 070 returns 496 / 481 / 421 parts for 2020 / 2010 / 2000, covering all 237 MORPC-lookup 070 geographies each year (~145 s per call).
+
+## 2026-09-24 — Treat 204 No Content as no rows; stop logging the API key (0.6.5)
+
+**204**: ACS 5-year 2017 lists `place/remainder (or part)` (070) in its geography.json but returns 204 No Content for every 070 request, in both ucgid and for/in forms. Since morpc 0.7.3 that raised `HTTPError` and stopped the whole fetch (and the pop-collect notebook run). `_fetch_variables` now catches `HTTPError` whose `response.status_code` is 204 (morpc ≥ 0.7.4 attaches the response), logs a warning, and treats that request as zero rows. If every request is empty, `CensusAPI.long` is an empty frame.
+
+**API key**: `geoinfo_from_params()` logged its params at INFO after adding `key` (#7, fixed in #8). morpc 0.7.5 redacts `key`/`token`/`api_key` in all `morpc.req` logs and errors (morpc/morpc-py#207); morpc-census now requires it.
+
+Tests: 3 in `tests/test_api.py` (204 chunk skipped, all-204 returns empty frame, other errors raise) and 1 in `tests/test_geos_hierarchical.py` (key not logged). 347 passing.
diff --git a/tests/test_api.py b/tests/test_api.py
index 7a23dfe..6fcfae6 100644
--- a/tests/test_api.py
+++ b/tests/test_api.py
@@ -1108,6 +1108,45 @@ def test_pseudo_ucgid_is_not_chunked(self):
         mock.assert_called_once()
         assert mock.call_args.kwargs['params']['ucgid'] == ucgid
 
+    @staticmethod
+    def _no_content():
+        from requests import HTTPError, Response
+        response = Response()
+        response.status_code = 204
+        return HTTPError("Request failed with status 204", response=response)
+
+    def test_no_content_chunk_is_treated_as_no_rows(self):
+        # The Census API answers 204 when it has no data for the requested geographies,
+        # e.g. ACS 5-year 2017 place/remainder parts.
+        api = self._make_api(1)
+        geoids = [f'0700000US39049{i:05d}99999' for i in range(150)]
+
+        def respond(url, params):
+            chunk = params['ucgid'].split(',')
+            if chunk[0] == geoids[100]:
+                raise self._no_content()
+            return [['GEO_ID', 'NAME'] + api.variables] + [[g, g] + ['1'] for g in chunk]
+
+        with patch('morpc.req.get_json_safely', side_effect=respond):
+            result = api._fetch_variables(api.request['url'], {'ucgid': ','.join(geoids)})
+        assert sorted(result['GEO_ID']) == geoids[:100]
+
+    def test_no_content_for_every_chunk_returns_empty_frame(self):
+        api = self._make_api(1)
+        with patch('morpc.req.get_json_safely', side_effect=self._no_content()):
+            result = api._fetch_variables(api.request['url'], {'ucgid': '0700000US390491800099999'})
+        assert result.empty
+        assert list(result.columns) == ['GEO_ID', 'NAME'] + api.variables
+
+    def test_other_http_errors_still_raise(self):
+        from requests import HTTPError, Response
+        api = self._make_api(1)
+        response = Response()
+        response.status_code = 400
+        with patch('morpc.req.get_json_safely', side_effect=HTTPError("bad", response=response)):
+            with pytest.raises(HTTPError):
+                api._fetch_variables(api.request['url'], {'ucgid': '0700000US390491800099999'})
+
 
 class TestFetchDispatch:
     """Tests for _fetch choosing between the group() and variable-list paths."""