The difficult nature of proving anything similar to this is astouding to me and missed in the article I was going over. They used (and the authors won the Netflix prize and was analyzing their data), also I should say using, Netflix to try to analyze the effects of this by their rating system. The giant flaw that I can readily see is the difference between adding new customers viewing the niche movies and existing customers moving from main stream movies into more niche movies. Further, they are missing the effect of the change in payment form for viewing the movies. Let me explain this last point a little more clearly...
To me as I read the article, they are confusing two differing payment systems that may lead to confusion in any business model derived from the data itself. Consider the difference between going to a movie store and using netflix. Where at a movie store you are paying for the usage of a single product over a single definite time period, while at the contrasting netflix, you are paying not for a single product over a definite period but rather the access to any product over any period (I do not use it but I am sure you cannot keep a movie indefinitely). When you blur the lines of product purchasing from a single product for the usage of any product, the consumer is likely to try differing product because the shift is from maximizing the utlity of a single purchase to maximizing the utility of the totality of what is offered. In other words, the shift would be from a Bertrand maximization to a Cournot maximization.
To simplify the last statement of the prior paragraph, consider the difference between going to a buffet and going to a restaurant. At most restaurants you pay for a specific meal of your choice. In this instance you are trying to maximize the effectiveness of the money are spending. In this instance, you are more likely to spend on what you know to be good or what you like more often than not or even further, what you have heard is more effective as opposed to what you do not know. In the opposition you have a buffet where rather than spending your money on a specific item, you are spending your money for the opportunity to try all the goods present at the buffet. In these two examples you are using two different standard normal distributions. One is going to have a higher middle while the other is going to have fatter tails. This does not necessarily desribe the long tail effect in which the consumer moves more towards the tails rather than stick in the middle of the distribution but rather a switch in the maximization game that the consumer is using to maximizing the experience.
Further on this, with the relative newness of the internut (intentional as a joke), you are introducing new customers and allowing some customers to a new environment. In order to prove the long tail effect, you would be viewing actual customers who were disposed to the same stimuli over and over again and switched preferences as a result of time and access an increased amount of goods present, however this is not the case in the example of Netflix versus the rest of the world of renting movies but rather what you are doing is switching business models and maximization models from one to another and only a buisness school student would miss this simple fact and say that the switch is from the long tail effect.
The effect described is not really a long tail effect but rather the natural effect of having equal access to all goods and naturally wanting to try more of them. In the model of netflix, the more you rent the more value you get out of your membership. Since there are a limited number of "bit hit" movies to view, the consumer is going to naturally try to view more on the tails of the spectrum. The long tail effect would be negated by this fact because essentially what it is describing is a wholesale movement from the center of a distribution to the tails of the distribution. In order to study this, one would have to use a controlled business model not differing two different models as you are simply comparing apples to oranges.
A long tail effect also would naturally be present as a person ages and tastes differ and the actual person becomes more sophisticated in their wants and beocme more involved in their passions. As I have, the access and lowered cost of the music has lead to more differentiation in my tastes of music. While I was young and less sophisticated, I would only listen to what was presented on the radio which is, in reality, garbage. My tastes evolved as I had access to much more music and was able to preview music at my convenience rather than buying and then experiencing it and risking losing the money on crap. If this is the long tail effect as understood by the authors of the Wharton School project, then they are just studying differing tastes and differing markets. In order to show a long tail effect you would have to isolate on the further sophistication of the audience viewing (or listening, or whatever manor) to show this effect and that is missed in the study.