(FORTUNE Magazine) -- Need proof that the CEO pay system is broken? Just look at the chasm between the bosses' flush fortunes and the plight of their shareholders.

The median pay for S&P 500 CEOs hit $8.4 million in 2005, according to Equilar, a California compensation-research firm - a raise from the heady days of 2001. But over the same period shareholders' returns were close to zero.

Misguided pay is expensive; worse, it is a recipe for decline, because it rewards self- defeating behavior, such as inflating earnings per share to wow Wall Street.

It is up to corporate boards to stand up to the bullying of the compensation consultants and to design plans that work for the best interests of the company. One place to start is to limit severance pay and eliminate bonuses when executives don't perform.

"You don't get paid for failure if things don't work out in a leveraged buyout," notes a prominent hedge fund manager. "Why should you in a public company?" And if performance is not sustained, CEOs should pay back the big bonus.

Such clawback mechanisms make it less likely that CEOs will "pump and dump" - that is, buy earnings through bad management (e.g., cutting R&D), then sell when the stock price begins to drop a few months later.

The principle at work should be to give the biggest rewards to those CEOs who leap the highest hurdles to build a competitive advantage for their companies.

CEOs need to be encouraged to act like owner-entrepreneurs who relish hatching pioneering products and smashing entrenched, wasteful practices. To get there, FORTUNE humbly proposes five commandments for CEO pay.

Most bonus plans are based on accounting earnings. But that number does not reflect the true profit that flows to owners. To generate earnings, CEOs use capital provided by shareholders, chiefly retained earnings.

Under accounting rules, companies pay no charge for that capital. But investors need a return on their capital, just as lenders charge interest for providing debt. Shareholders want the same return they'd receive from investing in any other stock with comparable risk. Today the number, called the cost of capital, is typically around 10%.

Until CEOs hand investors the cost of capital, they are not really making money for them. "The CEO should 'own' the improvement in profits he generates, but only after shareholders get their share first," concludes Bennett Stewart, co-founder of Stern Stewart & Co, a financial consulting firm.

A typical program pays a $5 million target bonus if the CEO raises earnings per share at a decent clip. But the CEO can get there the wrong way, by lavishing tons of "free" shareholder capital - chiefly retained earnings - on plants and products that yield low returns, or by borrowing money and earning a bit more than the 5% or 6% interest rate. Earnings per share keep climbing, the company keeps growing - and the stock price stagnates.

In a sound bonus plan, CEOs get big money only if they generate returns that exceed the cost of capital.

Says John Shiely, CEO of Briggs & Stratton, the engine and lawnmower manufacturer: "If you want CEOs to view capital the way shareholders do, you have to charge them for it."

So it's essential to deduct a charge for equity capital from accounting earnings to see how much money CEOs are really making for shareholders. That number is known as "economic profit," or "economic value added," a term coined by Stern Stewart. The key is to base bonuses on the returns CEOs garner on the new capital they get to use every year. That's what drives share prices.

The best plans give CEOs a fixed share of the economic profit they generate - the earnings they're reinvesting in new plants and products. CEOs who earn only the 10% cost of capital collect only the target bonus. Those CEOs who generate a higher return than the market is expecting, say 12% or 15%, should get a piece of the extra profit that is bound to drive up the share price.

Shiely at Briggs & Stratton receives a big bonus only from continuous improvements in profits - after paying for the new capital he uses. Headwaters, the construction-materials maker, and Herman Miller, the office-furniture manufacturer, also pay bonuses based on EVA. At DuPont, executives get bonuses based on earnings above a 6% return on capital.

The hallmark of a courageous, shareholder-friendly board is one that resists the lemming-like shift to restricted stock grants. In 2005 the S&P 500 companies increased such awards by 44%, to $15.3 billion, while stock option grants dropped 26%, to $30 billion, according to Jack Ciesielski, who writes a newsletter called The Analyst's Accounting Observer.

Just two years ago, most companies paid 70% of their long-term equity compensation in options and just 30% in stock. In part because options must now be expensed, that proportion is flip-flopping. The problem is that stock awards provide less power for the dollar than option grants.

As a rule of thumb, one share carries the same expense as three options. Hence, for every dollar CEOs raise the share price, they make three times as much from options as from a stock grant that costs the same. With options, CEOs have a lot more incentive to raise the stock price, because it magnifies their wealth faster.

To make matters worse, CEOs can still get rich if the stock price barely increases, or even if it falls. Ray Irani of Occidental Petroleum just got an award of $30.9 million that vests over three to five years. If the stock says flat, he's guaranteed almost $31 million.

Many companies are addressing the "pay for pulse" issue by requiring that CEOs meet targets for earnings per share or return on equity. In general it doesn't work.

"Most plans are designed to give the target grants if performance is average," says George Paulin, CEO of Frederic W. Cook & Co. "Boards usually get sandbagged on the goals."

CEOs love stock grants "because the payout is a lot safer," says Ira Kay, a head of the compensation practice at Watson Wyatt. "They have seen options go underwater and don't want to repeat the experience."

Put simply, America is swapping pay that was excessive but still risky for pay that's just as excessive but a sure thing.

The CEOs' success in enriching shareholders depends on a single, overriding factor: how much wealth the CEOs have tied up in stock and options and how fast that wealth will increase if they boost the stock price.

That's the conclusion of Steve O'Byrne of Shareholder Value Advisors, a compensation consulting firm. O'Byrne calls that magic factor "wealth leverage." And nothing juices it like stock options.

Still, conventional options have a big weakness. The exercise or "strike" price is the market price on the day of the grant, and it is fixed for ten years. So the CEO benefits from any increase in the stock price, even if it's far smaller than what's needed to give the shareholder a decent return on investment.

With a big enough grant, CEOs can get rich just by plowing back earnings, which does nothing for shareholders - or for the future of the company.

Today most option plans reward failure and punish success. Most boards are obsessed with giving CEOs a package comparable with that of their competitors. So if the stock price falls 20% this year, the board typically gives the CEO 20% more shares next year, to keep the dollar value stable. If the price jumps 20%, the grant is cut by one-fifth. The result, says O'Byrne, is that "you overpay unperforming managers, who stay on, and underpay the stars, who usually leave."

O'Byrne recommends that boards instead give CEOs equal annual grants of a fixed number of shares, for up to five years. So every time the CEO raises the stock price, the value of the future grants jumps; that is what wealth leverage is all about.

But the fixed-share idea works best if the board also subtracts the gains that come either from a general increase in the market or from extraordinary factors in the industry, such as the recent rise in oil prices. That way investors get the full benefit from the overall run-up in stocks, and CEOs gain only from their own contribution to shareholder wealth.

For example, in 2005, the board at Charles Schwab granted CEO Chuck Schwab three tranches of 800,000 options each. The three awards carry strike prices that are 12%, 25%, and 40% higher than the share price at the time of the grant.

"It was my idea," says Schwab. "I think it's right that the shareholders who put up the capital get a 12% preferred return before I benefit." What does Schwab think of the mania for restricted stock? He dismisses it as a "freebie."

A rare victory in the battle for more sensible compensation is that boards are finally insisting on vesting periods longer than the traditional three years. At Exxon Mobil top executives can't sell half of their restricted stock grants for five years, and must hold the other half for ten years or until retirement, whichever is later.

If boards insist on granting shares, that's the way to do it. Why? Because when the private fortunes of CEOs depend on raising the share price over the long term, they'll find a way.

In another positive trend, most companies now make CEOs hold anywhere from five to eight times their salary in stock. That forces CEOs to keep a big chunk of their wealth riding on the company's share price. At J.P. Morgan Chase, CEO Jamie Dimon is making senior executives retain 75% of all stock awards (after taxes and exercise costs) until they leave the company.

At Bristol-Myers Squibb, the CEO must hold at least 75% of newly vested stock or option grants for a minimum of two years. Right now CEO Peter Dolan has more than $140 million in stock and options at face value, most of which he'll have to hold for a long time. The future of his fortune is harnessed to the fortunes of his shareholders. That's just where a CEO should be.