diff --git a/benchmark/benchmark-pandoc.hs b/benchmark/benchmark-pandoc.hs
index 728e45b56..015d0135a 100644
--- a/benchmark/benchmark-pandoc.hs
+++ b/benchmark/benchmark-pandoc.hs
@@ -1,5 +1,5 @@
 import Text.Pandoc
-import Text.Pandoc.Shared (readDataFile, normalize)
+import Text.Pandoc.Shared (readDataFileUTF8, normalize)
 import Criterion.Main
 import Criterion.Config
 import Text.JSON.Generic
@@ -35,8 +35,8 @@ main :: IO ()
 main = do
   args <- getArgs
   (conf,_) <- parseArgs defaultConfig{ cfgSamples = Last $ Just 20 }  defaultOptions args
-  inp <- readDataFile (Just ".") "README"
-  inp2 <- readDataFile (Just ".") "tests/testsuite.txt"
+  inp  <- readDataFileUTF8 (Just ".") "README"
+  inp2 <- readDataFileUTF8 (Just ".") "tests/testsuite.txt"
   let opts = def{ readerSmart = True }
   let doc = readMarkdown opts $ inp ++ unlines (drop 3 $ lines inp2)
   let readerBs = map (readerBench doc) readers
diff --git a/pandoc.cabal b/pandoc.cabal
index 40d76be3f..8b40344cf 100644
--- a/pandoc.cabal
+++ b/pandoc.cabal
@@ -187,6 +187,9 @@ Source-repository head
 Flag blaze_html_0_5
   Description:   Use blaze-html 0.5 and blaze-markup 0.5
   Default:       True
+Flag embed_data_files
+  Description:   Embed data files in binary for relocatable executable.
+  Default:       False
 
 Library
   Build-Depends: base >= 4.2 && <5,
@@ -224,6 +227,10 @@ Library
   else
     build-depends:
                  blaze-html >= 0.4.3.0 && < 0.5
+  if flag(embed_data_files)
+     build-depends: file-embed >= 0.0.4 && < 0.1,
+                    template-haskell >= 2.4 && < 2.9
+     cpp-options:   -DEMBED_DATA_FILES
   if impl(ghc >= 7.0.1)
     Ghc-Options:   -O2 -rtsopts -Wall -fno-warn-unused-do-bind -dno-debug-output
   else
diff --git a/pandoc.hs b/pandoc.hs
index 7268f57f8..2300076c2 100644
--- a/pandoc.hs
+++ b/pandoc.hs
@@ -33,7 +33,7 @@ module Main where
 import Text.Pandoc
 import Text.Pandoc.PDF (tex2pdf)
 import Text.Pandoc.Readers.LaTeX (handleIncludes)
-import Text.Pandoc.Shared ( tabFilter, readDataFile, safeRead,
+import Text.Pandoc.Shared ( tabFilter, readDataFileUTF8, safeRead,
                             headerShift, normalize, err, warn )
 import Text.Pandoc.XML ( toEntities, fromEntities )
 import Text.Pandoc.SelfContained ( makeSelfContained )
@@ -889,24 +889,27 @@ main = do
                            E.catch (UTF8.readFile tp')
                              (\e -> if isDoesNotExistError e
                                        then E.catch
-                                             (readDataFile datadir $
-                                               "templates" </> tp')
+                                             (readDataFileUTF8 datadir
+                                                ("templates" </> tp'))
                                              (\e' -> let _ = (e' :: E.SomeException)
                                                      in throwIO e')
                                        else throwIO e)
 
   variables' <- case mathMethod of
                       LaTeXMathML Nothing -> do
-                         s <- readDataFile datadir $ "data" </> "LaTeXMathML.js"
+                         s <- readDataFileUTF8 datadir
+                                 ("data" </> "LaTeXMathML.js")
                          return $ ("mathml-script", s) : variables
                       MathML Nothing -> do
-                         s <- readDataFile datadir $ "data"</>"MathMLinHTML.js"
+                         s <- readDataFileUTF8 datadir
+                                 ("data"</>"MathMLinHTML.js")
                          return $ ("mathml-script", s) : variables
                       _ -> return variables
 
   variables'' <- if "dzslides" `isPrefixOf` writerName'
                     then do
-                        dztempl <- readDataFile datadir $ "dzslides" </> "template.html"
+                        dztempl <- readDataFileUTF8 datadir
+                                     ("dzslides" </> "template.html")
                         let dzcore = unlines $ dropWhile (not . isPrefixOf "<!-- {{{{ dzslides core")
                                              $ lines dztempl
                         return $ ("dzslides-core", dzcore) : variables'
@@ -927,15 +930,16 @@ main = do
               then do
                 csl <- CSL.parseCSL =<<
                         case mbCsl of
-                            Nothing      -> readDataFile datadir "default.csl"
+                            Nothing      -> readDataFileUTF8 datadir
+                                              "default.csl"
                             Just cslfile -> do
                                   exists <- doesFileExist cslfile
                                   if exists
                                      then UTF8.readFile cslfile
                                      else do
                                        csldir <- getAppUserDataDirectory "csl"
-                                       readDataFile (Just csldir)
-                                          (replaceExtension cslfile "csl")
+                                       readDataFileUTF8 (Just csldir)
+                                           (replaceExtension cslfile "csl")
                 abbrevs <- maybe (return []) CSL.readJsonAbbrevFile cslabbrevs
                 return $ Just csl { CSL.styleAbbrevs = abbrevs }
               else return Nothing
diff --git a/src/Text/Pandoc/SelfContained.hs b/src/Text/Pandoc/SelfContained.hs
index e468d504d..4b52a6f13 100644
--- a/src/Text/Pandoc/SelfContained.hs
+++ b/src/Text/Pandoc/SelfContained.hs
@@ -41,7 +41,7 @@ import System.FilePath (takeExtension, dropExtension, takeDirectory, (</>))
 import Data.Char (toLower, isAscii, isAlphaNum)
 import Codec.Compression.GZip as Gzip
 import qualified Data.ByteString.Lazy as L
-import Text.Pandoc.Shared (findDataFile, renderTags')
+import Text.Pandoc.Shared (readDataFile, renderTags')
 import Text.Pandoc.MIME (getMimeType)
 import System.Directory (doesFileExist)
 import Text.Pandoc.UTF8 (toString,  fromString)
@@ -55,18 +55,8 @@ getItem userdata f =
                       ".gz" -> getMimeType $ dropExtension f
                       x     -> getMimeType x
        exists <- doesFileExist f
-       if exists
-          then do
-            cont <- B.readFile f
-            return (cont, mime)
-          else do
-            res <- findDataFile userdata f
-            exists' <- doesFileExist res
-            if exists'
-               then do
-                 cont <- B.readFile res
-                 return (cont, mime)
-               else error $ "Could not find `" ++ f ++ "'"
+       cont <- if exists then B.readFile f else readDataFile userdata f
+       return (cont, mime)
 
 -- TODO - have this return mime type too - then it can work for google
 -- chart API, e.g.
diff --git a/src/Text/Pandoc/Shared.hs b/src/Text/Pandoc/Shared.hs
index dee10cf9b..71bdca9dd 100644
--- a/src/Text/Pandoc/Shared.hs
+++ b/src/Text/Pandoc/Shared.hs
@@ -1,4 +1,4 @@
-{-# LANGUAGE DeriveDataTypeable #-}
+{-# LANGUAGE DeriveDataTypeable, CPP #-}
 {-
 Copyright (C) 2006-2010 John MacFarlane <jgm@berkeley.edu>
 
@@ -64,8 +64,8 @@ module Text.Pandoc.Shared (
                      renderTags',
                      -- * File handling
                      inDirectory,
-                     findDataFile,
                      readDataFile,
+                     readDataFileUTF8,
                      -- * Error handling
                      err,
                      warn,
@@ -89,13 +89,18 @@ import System.FilePath ( (</>) )
 import Data.Generics (Typeable, Data)
 import qualified Control.Monad.State as S
 import Control.Monad (msum)
-import Paths_pandoc (getDataFileName)
 import Text.Pandoc.Pretty (charWidth)
 import System.Locale (defaultTimeLocale)
 import Data.Time
 import System.IO (stderr)
 import Text.HTML.TagSoup (renderTagsOptions, RenderOptions(..), Tag(..),
          renderOptions)
+import qualified Data.ByteString as B
+#ifdef EMBED_DATA_FILES
+import Data.FileEmbed
+#else
+import Paths_pandoc (getDataFileName)
+#endif
 
 --
 -- List processing
@@ -499,20 +504,28 @@ inDirectory path action = do
   setCurrentDirectory oldDir
   return result
 
--- | Get file path for data file, either from specified user data directory,
--- or, if not found there, from Cabal data directory.
-findDataFile :: Maybe FilePath -> FilePath -> IO FilePath
-findDataFile Nothing f = getDataFileName f
-findDataFile (Just u) f = do
-  ex <- doesFileExist (u </> f)
-  if ex
-     then return (u </> f)
-     else getDataFileName f
+readDefaultDataFile :: FilePath -> IO B.ByteString
+readDefaultDataFile fname =
+#ifdef EMBED_DATA_FILES
+  TODO
+#else
+  getDataFileName fname >>= B.readFile
+#endif
 
 -- | Read file from specified user data directory or, if not found there, from
 -- Cabal data directory.
-readDataFile :: Maybe FilePath -> FilePath -> IO String
-readDataFile userDir fname = findDataFile userDir fname >>= UTF8.readFile
+readDataFile :: Maybe FilePath -> FilePath -> IO B.ByteString
+readDataFile Nothing fname = readDefaultDataFile fname
+readDataFile (Just userDir) fname = do
+  exists <- doesFileExist (userDir </> fname)
+  if exists
+     then B.readFile (userDir </> fname)
+     else readDefaultDataFile fname
+
+-- | Same as 'readDataFile' but returns a String instead of a ByteString.
+readDataFileUTF8 :: Maybe FilePath -> FilePath -> IO String
+readDataFileUTF8 userDir fname =
+  UTF8.toString `fmap` readDataFile userDir fname
 
 --
 -- Error reporting
diff --git a/src/Text/Pandoc/Templates.hs b/src/Text/Pandoc/Templates.hs
index 75c133101..bbdb4adc4 100644
--- a/src/Text/Pandoc/Templates.hs
+++ b/src/Text/Pandoc/Templates.hs
@@ -80,7 +80,7 @@ import Text.Blaze (preEscapedString, Html)
 #endif
 import Text.Pandoc.UTF8 (fromStringLazy)
 import Data.ByteString.Lazy (ByteString)
-import Text.Pandoc.Shared (readDataFile)
+import Text.Pandoc.Shared (readDataFileUTF8)
 import qualified Control.Exception.Extensible as E (try, IOException)
 
 -- | Get default template for the specified writer.
@@ -98,7 +98,7 @@ getDefaultTemplate user writer = do
        "multimarkdown"   -> getDefaultTemplate user "markdown"
        "markdown_github" -> getDefaultTemplate user "markdown"
        _        -> let fname = "templates" </> "default" <.> format
-                   in  E.try $ readDataFile user fname
+                   in  E.try $ readDataFileUTF8 user fname
 
 data TemplateState = TemplateState Int [(String,String)]
 
diff --git a/src/Text/Pandoc/Writers/Docx.hs b/src/Text/Pandoc/Writers/Docx.hs
index b0f31ac45..706ced967 100644
--- a/src/Text/Pandoc/Writers/Docx.hs
+++ b/src/Text/Pandoc/Writers/Docx.hs
@@ -29,14 +29,12 @@ Conversion of 'Pandoc' documents to docx.
 -}
 module Text.Pandoc.Writers.Docx ( writeDocx ) where
 import Data.List ( intercalate )
-import System.FilePath ( (</>) )
 import qualified Data.ByteString.Lazy as B
 import qualified Data.Map as M
 import qualified Text.Pandoc.UTF8 as UTF8
 import System.IO ( stderr )
 import Codec.Archive.Zip
 import Data.Time.Clock.POSIX
-import Paths_pandoc ( getDataFileName )
 import Text.Pandoc.Definition
 import Text.Pandoc.Generic
 import System.Directory
@@ -104,15 +102,8 @@ writeDocx opts doc@(Pandoc (Meta tit auths date) _) = do
   refArchive <- liftM toArchive $
        case writerReferenceDocx opts of
              Just f -> B.readFile f
-             Nothing -> do
-               let defaultDocx = getDataFileName "reference.docx" >>= B.readFile
-               case datadir of
-                     Nothing  -> defaultDocx
-                     Just d   -> do
-                        exists <- doesFileExist (d </> "reference.docx")
-                        if exists
-                           then B.readFile (d </> "reference.docx")
-                           else defaultDocx
+             Nothing -> (B.fromChunks . (:[])) `fmap`
+                          readDataFile datadir "reference.docx"
 
   (newContents, st) <- runStateT (writeOpenXML opts{writerWrapText = False} doc)
                        defaultWriterState
diff --git a/src/Text/Pandoc/Writers/EPUB.hs b/src/Text/Pandoc/Writers/EPUB.hs
index e7feace7a..4ffff0054 100644
--- a/src/Text/Pandoc/Writers/EPUB.hs
+++ b/src/Text/Pandoc/Writers/EPUB.hs
@@ -30,13 +30,13 @@ Conversion of 'Pandoc' documents to EPUB.
 module Text.Pandoc.Writers.EPUB ( writeEPUB2, writeEPUB3 ) where
 import Data.IORef
 import Data.Maybe ( fromMaybe, isNothing )
-import Data.List ( isPrefixOf, isInfixOf, intercalate )
+import Data.List ( isInfixOf, intercalate )
 import System.Environment ( getEnv )
 import Text.Printf (printf)
 import System.FilePath ( (</>), takeBaseName, takeExtension, takeFileName )
 import qualified Data.ByteString.Lazy as B
 import qualified Data.ByteString.Lazy.Char8 as B8
-import Text.Pandoc.UTF8 ( fromStringLazy )
+import Text.Pandoc.UTF8 ( fromStringLazy, toString )
 import Codec.Archive.Zip
 import Data.Time.Clock.POSIX
 import Data.Time
@@ -321,7 +321,8 @@ writeEPUB version opts doc@(Pandoc meta _) = do
   -- stylesheet
   stylesheet <- case writerEpubStylesheet opts of
                    Just s  -> return s
-                   Nothing -> readDataFile (writerUserDataDir opts) "epub.css"
+                   Nothing -> toString `fmap`
+                              readDataFile (writerUserDataDir opts) "epub.css"
   let stylesheetEntry = mkEntry "stylesheet.css" $ fromStringLazy stylesheet
 
   -- construct archive
diff --git a/src/Text/Pandoc/Writers/ODT.hs b/src/Text/Pandoc/Writers/ODT.hs
index cbff88be5..5e92cfcec 100644
--- a/src/Text/Pandoc/Writers/ODT.hs
+++ b/src/Text/Pandoc/Writers/ODT.hs
@@ -35,15 +35,13 @@ import qualified Data.ByteString.Lazy as B
 import Text.Pandoc.UTF8 ( fromStringLazy )
 import Codec.Archive.Zip
 import Data.Time.Clock.POSIX
-import Paths_pandoc ( getDataFileName )
 import Text.Pandoc.Options ( WriterOptions(..) )
-import Text.Pandoc.Shared ( stringify )
+import Text.Pandoc.Shared ( stringify, readDataFile )
 import Text.Pandoc.ImageSize ( readImageSize, sizeInPoints )
 import Text.Pandoc.MIME ( getMimeType )
 import Text.Pandoc.Definition
 import Text.Pandoc.Generic
 import Text.Pandoc.Writers.OpenDocument ( writeOpenDocument )
-import System.Directory
 import Control.Monad (liftM)
 import Network.URI ( unEscapeString )
 import Text.Pandoc.XML
@@ -59,15 +57,8 @@ writeODT opts doc@(Pandoc (Meta title _ _) _) = do
   refArchive <- liftM toArchive $
        case writerReferenceODT opts of
              Just f -> B.readFile f
-             Nothing -> do
-               let defaultODT = getDataFileName "reference.odt" >>= B.readFile
-               case datadir of
-                     Nothing  -> defaultODT
-                     Just d   -> do
-                        exists <- doesFileExist (d </> "reference.odt")
-                        if exists
-                           then B.readFile (d </> "reference.odt")
-                           else defaultODT
+             Nothing -> (B.fromChunks . (:[])) `fmap`
+                           readDataFile datadir "reference.odt"
   -- handle pictures
   picEntriesRef <- newIORef ([] :: [Entry])
   let sourceDir = writerSourceDirectory opts