For the record, legitimate case when many empty dicts are created, and few are populated, is the collections-free approach to defaultdict(dict):

    mydict.setdefault(key1, {})[key2] = val

For, say, 100 unique key1s, and 10,000 total key1/key2 pairs, you'd create 10,000 empty dicts, discarding 9,900 of them. Granted, collections.defaultdict(dict) is even better (avoids the 9,900 unused dicts entirely), but I see the setdefault pattern enough, usually with list or dict, that it's not totally unreasonable to account for it.