How do I correctly insert unicode in an HTML title using JavaScript?

I'm seeing some weird behavior when I'm setting the title of an HTML page using JavaScript. If I insert html character references directly into the title the Unicode renders correctly, for instance:

<title>&#21543;&#20986;</title>

But if I attempt to use html characters references via JavaScript, something seems to be converting the & to (& amp ;) (separating them so SO doesn't just turn it back into ampersand) and thus breaking the encoding, causing it to be rendered as the full coded string:

function execTitleChange() {
  document.title = "吧出";
}

(I should note that this is a little bit of speculation; when I introspect the DOM using Firebug after executing this JavaScript function, that's where I see the & instead of &.)

If I use \u encoded Unicode characters when setting the value from JavaScript then everything works correctly again:

function execTitleChange() {
  document.title = "\u5427\u51fa";
}

The fact that \u encoded characters work kind of makes sense to me since I think that's how JavaScript represents Unicode characters but I'm stumped as to why the behavior would be different when using the html character references.