I have written extensively about perfect competition and the fact that the claims associated with it cannot be proven scientifically. In fact, computer simulations of perfect competition show conclusively the same price and amounts as in a monopoly. I've also written about Joseph Stiglitz' explanation of why this might be a logical outcome of the theory.
There is a further point that I want to make, and that Hill and Myatt make in their excellent book The Economics Anti-Textbook. It is the point about the "typical cost curve" that underlies the theory of perfect competition.
For the theory to produce the results that economists claim, firms must be so small that they don't have market power: any change in production quota is too small as to affect market prices. The chief thing that guarantees their small sizes is the typical u-shaped average cost curve, because profit-maximising firms will stop producing when marginal cost of production equals the marginal revenue from sales. The problem is, that the typical u-shaped average cost curve is a total fabrication pulled out of thin air. What economists call "typical" is known at least since 1952 to have no factual base whatsoever.
In a survey by Wilford J. Eiteman and Glenn E. Guthrie in 1952 (published in The Shape of the Average Cost Curve, American Economic Review, 42.5: 832–838) managers of 334 companies were shown a number of different cost curves, and asked to
specify which one best represented the company’s cost curve. A stunning 95% of managers responding to the survey chose cost curves with constant or falling costs. That leaves at most 5% for the "typical" u-shape.
A. S. Blinder, former vice chairman of the American Economics Association, conducted the same type of survey in 1998, which involved 200 US firms in a
sample that should be representative of the US economy at large.
He found that about 40% of firms reported falling variable or
marginal cost, and 48.4% reported constant marginal/variable cost
(Asking about Prices: A New Approach to Understanding Price Stickiness. Russell Sage Foundation, New York). A minority of about 11% of those surveyed reported the typical u-shape.
What does all this tell us? Economists know at least for 60 (that's sixty) years that the tale of u-shaped cost curves -- and as a consequence the majesty of perfect competition that rests heavily on this claim -- is a falsehood, yet it's still the staple of every economics textbook. Note that also the Wikipedia page linked above did not mention the fact that the claim has been empirically proven to be false, until I added it on January 3, 2015 (see page history).
The u-shaped cost curve is not a simplification either, because assuming a constant average cost would be much simpler. But then, constant average cost does not guarantee the firms to remain small, and hence perfect competition to work.
No, the typical u-shaped cost curve is a complete fabrication pulled out of thin air in order for the theory of perfect competition to yield the results that economists wish it to have but cannot produce otherwise. After all, who wants to know a truth that contradicts the dogma of perfect competition?
Now, how's that for a value-free and results-open science!
Update [28.12.2015]: I just noticed that the Wikipedia page that I edited (see above) has been modified to begin with the statement: "Some argue that cost curves are not typically U-shaped." If nothing else, this shows how insidiously economists misrepresent facts. Indeed, the statement "some argue that..." makes it seem as if the not-U-shaped cost curves are a position held by a minority of economists ("some") when what follows is actually real evidence. Not just "theory", but observations in the real world. Something that "some" people call science. I modified the article to state "Evidence shows that cost curves are not typically U-shaped." Let's see how long this modification lasts.