diff --git a/babel/messages/jslexer.py b/babel/messages/jslexer.py
index c1c25575c..1264b2dbc 100644
--- a/babel/messages/jslexer.py
+++ b/babel/messages/jslexer.py
@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 """
     babel.messages.jslexer
     ~~~~~~~~~~~~~~~~~~~~~~
@@ -27,6 +28,7 @@
 line_re = re.compile(r'(\r\n|\n|\r)')
 line_join_re = re.compile(r'\\' + line_re.pattern)
 uni_escape_re = re.compile(r'[a-fA-F0-9]{1,4}')
+hex_escape_re = re.compile(r'[a-fA-F0-9]{1,2}')
 
 Token = namedtuple('Token', 'type value lineno')
 
@@ -127,6 +129,17 @@ def unquote_string(string):
             else:
                 add(next_char)
 
+        # hex escapes. conversion from 2-digits hex to char is infallible
+        elif next_char in 'xX':
+            escaped = hex_escape_re.match(string, escape_pos + 2)
+            if escaped is not None:
+                escaped_value = escaped.group()
+                add(chr(int(escaped_value, 16)))
+                pos = escape_pos + 2 + len(escaped_value)
+                continue
+            else:
+                add(next_char)
+
         # bogus escape.  Just remove the backslash.
         else:
             add(next_char)
diff --git a/tests/messages/test_jslexer.py b/tests/messages/test_jslexer.py
index 35204eee0..bd6322e9e 100644
--- a/tests/messages/test_jslexer.py
+++ b/tests/messages/test_jslexer.py
@@ -4,6 +4,8 @@
 def test_unquote():
     assert jslexer.unquote_string('""') == ''
     assert jslexer.unquote_string(r'"h\u00ebllo"') == u"hëllo"
+    assert jslexer.unquote_string(r'"h\xebllo"') == u"hëllo"
+    assert jslexer.unquote_string(r'"\xebb"') == u"ëb"
 
 
 def test_dollar_in_identifier():