If managers free to engage in insider trading know that the next piece of news to be released will cause the stock price to rise, they will buy. If they know that the next piece of news to be released will cause the stock price to fall, they will sell and then buy back later. They don't care whether the news is good or bad--either way they will profit, and either way they will profit equally.
What the ability to engage in insider trading does is that it gives managers an incentive to make the price of the stock vary--they don't care which way. Thus it cannot "serve a useful purpose as an executive compensation device" and cannot "motivate managers to maximize the value of the firm" to shareholders.
Insider trading makes executives' portfolios' long not the company but long the volatility of the company. And shareholders don't want executives making decisions that make the value of companies they own more volatile: stock market investments are risky enough as it is without giving executives reasons to boost the volatility pot.
I think that this really is one of the hardest things about modern economics. The idea of analyzing behavior based on what the incentives are is atremendously powerful tool. The problem is that it is possible to mistake the incentives involved and reach a very poor conclusion.
The part that is scary is that he is actually responding to an actual argument suggesting that permitting insider trading might be a way of incenting executives to maximize the value of a firm.
I am starting to suspect that the analogy of a market as being a structured competition (which I first picked up from reading Joseph Heath) is actually a very useful analogy.