avoid .net XmlTextWriter escaping ampersand in front of character reference

Can someone please be explain why System.Xml.XmlTextWriter escapes the ampersand in front of what I thought was a valid unicode character reference:

[TestMethod]
public void TestEncoding()
{
    using (StringWriter sw = new StringWriter())
    {
        XmlWriter writer = new XmlTextWriter(sw);
        const string testChar = "–";
        writer.WriteString(testChar);
        Assert.AreEqual(testChar, sw.ToString());
        writer.Close();
    }
}

fails with the message

Assert.AreEqual failed. Expected:<&#8211;>. Actual:<&amp;#8211;>.

Is there a way to escape this behavior, while correctly maintaining conversion of isolated or invalid ampersands to & ?