diff --git a/ftfy/badness.py b/ftfy/badness.py
index 38ec1f4..47e56f7 100644
--- a/ftfy/badness.py
+++ b/ftfy/badness.py
@@ -365,6 +365,13 @@
     ^[ÃÂ][ ]
     |
 
+    # Windows-1252 mojibake of three-byte UTF-8 sequences in the U+2000
+    # punctuation and symbol blocks, whose lead byte 0xE2 shows up as 'â'.
+    # This is restricted to 'â' on purpose: the same shape with other
+    # accented letters is ambiguous, e.g. 'Brontë…”' is not mojibake.
+    â[{common}][{start_punctuation}{end_punctuation}{currency}{numeric}{common}]
+    |
+
     # Cases where  precedes a character as an encoding of exactly the same
     # character, and the character is common enough
     [a-z.,?!{end_punctuation}] Â [ {start_punctuation}{end_punctuation}]
diff --git a/tests/test-cases/synthetic.json b/tests/test-cases/synthetic.json
index a939311..b232be7 100644
--- a/tests/test-cases/synthetic.json
+++ b/tests/test-cases/synthetic.json
@@ -204,5 +204,12 @@
         "original": "OÙ ET QUAND?",
         "fixed": "OÙ ET QUAND?",
         "expect": "pass"
+    },
+    {
+        "label": "Synthetic: vulgar fractions in Windows-1252 mojibake",
+        "comment": "issue #233. The lead byte of U+2153 is 0xE2, which shows up as 'â'; the second byte lands in the 'common' category, which on its own is not evidence of mojibake.",
+        "original": "â…“ and â…›",
+        "fixed": "⅓ and ⅛",
+        "expect": "pass"
     }
 ]
\ No newline at end of file