Skip to main content
There’s no single API call that answers “which journals does institution X cite most?” — the referenced_works field on each work contains outgoing citations as raw IDs, not journal names. But with cursor paging and batching, you can build the full picture. We’ll use the Santa Fe Institute (496 works in 2024) as the example.

​
The approach

  1. Cursor through your institution’s works, collecting every referenced_works ID (with duplicates — if 5 papers cite the same work, that’s 5 references)
  2. Batch-fetch the unique IDs to build a work-to-journal lookup
  3. Count journals against the full reference list

​
Step 1: Collect referenced work IDs

Page through works with select=id,referenced_works to minimize payload: 
https://api.openalex.org/works?filter=authorships.institutions.id:I1308548392,publication_year:2024&select=id,referenced_works&per_page=100&cursor=*
Each work’s referenced_works is an array of OpenAlex IDs: 
{
  "id": "https://openalex.org/W4392028279",
  "referenced_works": [
    "https://openalex.org/W1583139675",
    "https://openalex.org/W2001771035",
    "https://openalex.org/W2025849425"
  ]
}
Follow meta.next_cursor until it returns null — for 496 works that’s 5 pages.

​
Step 2: Batch and count by journal

Batch-fetch the unique IDs with select=id,primary_location to build a lookup from work ID to journal: 
https://api.openalex.org/works?filter=openalex:W1583139675|W2001771035|W2025849425|...&select=id,primary_location&per_page=100
Then iterate over the full (non-deduplicated) reference list, look up each work’s journal, and count.

​
Full script

import requests
from collections import Counter

BASE = "https://api.openalex.org"
INST = "I1308548392"  # Santa Fe Institute
YEARS = "2024"

def api(endpoint, params):
    return requests.get(f"{BASE}/{endpoint}", params=params).json()

# Step 1: collect ALL referenced work IDs (keeping duplicates)
all_refs = []
cursor = "*"
while cursor:
    resp = api("works", {
        "filter": f"authorships.institutions.id:{INST},publication_year:{YEARS}",
        "select": "id,referenced_works",
        "per_page": 100,
        "cursor": cursor,
    })
    for work in resp["results"]:
        for ref in work.get("referenced_works", []):
            all_refs.append(ref.split("/")[-1])
    cursor = resp["meta"].get("next_cursor")

unique_refs = list(set(all_refs))
print(f"{len(all_refs)} total references, {len(unique_refs)} unique works")

# Step 2: build work → journal lookup (fetch unique IDs only)
work_to_journal = {}
for i in range(0, len(unique_refs), 100):
    batch = "|".join(unique_refs[i:i+100])
    results = api("works", {
        "filter": f"openalex:{batch}",
        "select": "id,primary_location",
        "per_page": 100,
    })["results"]
    for w in results:
        loc = w.get("primary_location") or {}
        source = (loc.get("source") or {}).get("display_name")
        if source:
            work_to_journal[w["id"].split("/")[-1]] = source

# Step 3: count journals against the FULL reference list (with duplicates)
journal_counts = Counter()
for ref_id in all_refs:
    journal = work_to_journal.get(ref_id)
    if journal:
        journal_counts[journal] += 1

print(f"\n{'Journal':<55} {'Refs':>5}")
print("-" * 62)
for journal, count in journal_counts.most_common(15):
    print(f"  {journal:<53} {count:>5}")

​
Example results

Top journals cited by Santa Fe Institute authors in 2024 (from a sample of references):
JournalReferences
Proceedings of the National Academy of Sciences14
Science13
PLoS ONE12
Nature7
Environmental Science & Technology4
Nature Plants3
Forest Ecology and Management3
Frontiers in Ecology and Evolution3
These counts are from a small sample — running the full script produces a complete ranking across all 496 works and their thousands of outgoing references. The interdisciplinary spread (PNAS, Science, Nature alongside ecology journals) is characteristic of SFI’s research.
This recipe requires many API calls — roughly one per 100 referenced works. For an institution with 500 works averaging 30 references each, expect ~150 calls. Add an api_key parameter and a brief time.sleep(0.1) between batches to stay within rate limits.