New Journalism, Old Journalism, and Still the Same Neglect of the Variance

While I’m not normally a huge fan of Gawker, this is a very informed critique of the kind of depths that the self-important “data journalism” has fallen to.

The “now-cast” of the sort that the article refers to, i.e. the “what if election were held now” type of “prediction” thrown about even by so-called data journalists like 538 and NYT Upshot, is totally meaningless outside of context.   Current polls may or may not show an upsurge of support for Trump around the GOP  convention, or not.  They may or may not show an upsurge of support for Clinton after the Democratic convention, or not.  But the real question is how much information they provide for how things might pan out by November.

The answer here needs to be much more nuanced than not:  the common practice is to ignore these “convention bumps” as meaningless–which they probably are if their role is only to help “predict” the election outcome.  But, as pointed out by Gelman, Rivers, and others, there’s something more going on there.  It is undeniable that there is an observable bump:  what is driving this bump?  Are people showing up who weren’t there?  Are people who were there not showing up?  Are people genuinely changing their mind?  Where is this variability coming from?  The variability of a given demographic, preferably one defined as precisely as possible, is critical:  variable voters are persuadable, and good campaigns aimed at persuadable voters stand far better chance of paying off than simply throwing money–if there are enough persuadable voters, given the right message and the messengers.

In context of the convention bumps, then, it is pointless to claim that the bump implies a greater chance of Trump win, or even that the bump is meaningless because it “will” go away–because we don’t know if it really will go away.  We need to understand where the bump is coming from if we want to take it seriously, or to dismiss it as a mirage.  Even then, of course, we wouldn’t know for sure–we can only understand the variances, with which we can form confidence intervals around.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s