diff --git a/dask/dataframe/io.py b/dask/dataframe/io.py
index 2b3cbf74c94..08d7a9891ad 100644
--- a/dask/dataframe/io.py
+++ b/dask/dataframe/io.py
@@ -12,7 +12,7 @@
 from itertools import count
 from operator import getitem
 
-from ..compatibility import StringIO, unicode, range
+from ..compatibility import BytesIO, unicode, range
 from ..utils import textblock
 from .. import array as da
 
@@ -21,12 +21,6 @@
 from .shuffle import set_partition
 
 
-def _StringIO(data):
-    if isinstance(data, bytes):
-        data = data.decode()
-    return StringIO(data)
-
-
 def file_size(fn, compression=None):
     """ Size of a file on disk
 
@@ -131,12 +125,12 @@ def read_csv(fn, *args, **kwargs):
 
     # Create dask graph
     name = 'read-csv' + next(tokens)
-    dsk = dict(((name, i), (rest_read_csv, (_StringIO,
+    dsk = dict(((name, i), (rest_read_csv, (BytesIO,
                                (textblock, fn,
                                    i*chunkbytes, (i+1) * chunkbytes,
                                    kwargs['compression']))))
                for i in range(1, nchunks))
-    dsk[(name, 0)] = (first_read_csv, (_StringIO,
+    dsk[(name, 0)] = (first_read_csv, (BytesIO,
                        (textblock, fn, 0, chunkbytes, kwargs['compression'])))
 
     result = DataFrame(dsk, name, columns, divisions)