If you go to a store you’ll find that a huge quantity of the goods on sale are labeled to indicate which brand makes them. There’s a good reason for this. Normally, how much you’re willing to pay for a good or service depends on the quality of the good or service in question. But there’s no way to sample the quality of a can of soda without buying it first. So how am I to know whether or not I want to buy that can of Diet Coke? Well it’s simple. I may not have had that can of Diet Coke before, but I have had many other cans of Diet Coke. And I can infer that the Coca-Cola corporation, having invested a great deal of time and money in building the Diet Coke band is going to make a good-faith effort to turn out a consistent product. That doesn’t necessarily mean everyone will like the taste of Diet Coke. But it does mean everyone knows more-or-less what Diet Coke tastes like, and then they can make their soda-consumption choices in a coherent way.
Politics, he argues, should be the same way. Political labels provide a coherent and efficient method of sorting candidates, making it easy for the electorate to identify which politicians match their policy preferences.
There's something to this, but I think it may assume a considerably more informed and engaged electorate than we have. The better analogy might not be to labels, but to one's favorite sports team. Most people come by their favorite teams fairly arbitrarily (most commonly, the location where they're born). And while theoretically the labeling of different teams could enable easy sorting for a fan to optimize her rooting pleasure ("The Devils specialize in tight, neutral-zone trap defense backed by stellar goaltending. That's my kind of hockey!"), nobody really does that.
Political participation, by and large, seems to operate the same way. The best predictor of what party a given American supports is what party their parents support. People support their political "team" because it's fun to root for a side (indeed, this is one of the ways to overcome widespread rational political ignorance), not because they've sampled the various brands and come to a rational voting decision.
Indeed, when one tries to break down the analogy, a host of problems crop up. All one needs to know about Diet Coke to know if one will like it or not is to buy one can and taste it. There's no easy parallel in politics. Political campaigns are highly suffused in rhetoric designed to appeal to all voters (everybody hates crime, everybody likes education, everybody wants more jobs), so there isn't really effective branding along a host of politically salient axes. The real impact of a given politician on any specific aspect of any particular citizen's life is going to be, at best, highly attenuated, making it difficult to say whether John Doe's advocacy of X policy really is something one likes or dislikes beyond mere gut instinct.
Of course, if one has extremely well-developed policy preferences and pays close attention to the political arena, then it is possible to rationally sort good politicians from bad ones. But if one is that engaged, then one likely doesn't need labels to serve as a proxy. Put another way, the level of participation one needs to have in the soda market in order to rationally order preferences is very low, while the level of participation one needs to have in the political system in order to rationally order preferences is extremely high. Political labels don't really overcome this information deficit so much as they give people a reason to participate in politics notwithstanding their general lack of information.