March 26 (Bloomberg) -- $12.1 million. That's the average total compensation in 2007 for a chief executive officer at the 107 largest U.S.-based companies who have so far released their proxy statements. The pay increase was 12 percent from 2006.

Pay ranged from a low of $175,000 for Warren Buffett of Berkshire Hathaway Inc. to the $56 million for Kenneth Chenault of American Express Co.

That's what my analysis shows of early reporting by companies with current market caps of at least $2 billion whose fiscal years ended Dec. 31, 2007. The companies disclosed two years of CEO pay in their proxy statements under the reporting system started in 2006 by the Securities and Exchange Commission. All the chiefs had been in their jobs at least two years. (For a list of the 107 executives and their companies and pay and performance data, click HERE.)

Four CEOS and their compensation stand out:

-- The $55.8 million in total pay for Chenault of New York- based American Express was an increase of 141 percent over his $23.1 million in 2006. The company's total return was negative 13 percent.

You may wonder what happened to Lloyd Blankfein, CEO of Goldman Sachs Group Inc., who earned $57 million in 2007. Goldman was excluded from this study because it used the old SEC format for 2006 and the new one for its fiscal year ended Nov. 30, 2007, making comparisons problematic.

-- For years Buffett of Omaha-based Berkshire Hathaway earned a salary of $100,000. Some miscellaneous compensation increased that to $175,000 in 2007. Total return was 29 percent. There is a strong but negative relationship between pay and performance if you compare American Express with Berkshire Hathaway.

Lucier's Increase

-- Gregory Lucier of San Diego area-based Invitrogen Corp., a biotech company, received the largest percentage increase in total pay in 2007: 522 percent, to $29 million from the year- earlier $4.7 million. Total return for the year was 65 percent. On March 1, 2007, the company's compensation committee awarded Lucier 405,000 free shares, to be earned over three years provided certain future share prices are obtained. At grant, the shares were worth $25.6 million. Lucier's committee said it didn't intend to grant him any further equity awards for a three-year period. It could, of course, have adopted a smoother alternative, giving him a third of the award each year for three years. That would have lowered significantly his 2007 pay increase.

O'Neil's Decrease

-- E. Stanley O'Neal of New York-based Merrill Lynch and Co. Inc. received the largest decrease in pay: 97 percent to $1.5 million from $48.5 million in 2006. This was one of the rare and clear-cut cases of pay for performance. According to the company's proxy statement, O'Neal ``retired'' on Oct. 30, 2007.

My analysis also shows:

-- The average total return for 2007 for the 107 companies was 10 percent versus 5.5 percent for the return on the Standard & Poor's 500 Index.

-- Total return performance ranged from the 299 percent for Sioux City, Iowa-based Terra Industries Inc., a nitrogen fertilizer producer, to negative 68 percent for Seattle-based Washington Mutual Inc., the largest U.S. savings and loan, which analysts have said is a takeover target because of losses on subprime mortgages.

-- The close relationship between the 12 percent average total compensation increase and the 10 percent average total return suggests there is pay-for-performance out there. Yet it is relatively weak, accounting for just 13 percent of the variation in pay increases. That means that 87 percent of what CEOs received in pay increases in 2007 had nothing to do with the returns they delivered to shareholders.

Pay and Performance

-- The return on 10-year U.S. Treasury bonds in 2007 was about 4.8 percent. By comparison, the CEO who received the average 12 percent pay increase produced only a 5.4 percent average return above the Treasury rate.

-- The 12 percent compensation increase, or $1.4 million, was more than triple the 3.6 percent increase in the national weekly wage for 2007, according to U.S. Bureau of Labor Statistics.

Finally, my analysis shows the relationship between the size of an increase for a CEO and the performance delivered to shareholders needs to be strengthened mightily.

Though it likely will never happen, the average pay increase for CEOs as a group should be lowered much closer to what the typical American gets, except when the average shareholder obtains a return that is substantially more, not just a bit more, than what he could have obtained taking no risk at all.

And the flip side, of course, is that when corporate results are poor, the average pay increase for a CEO should be lower than that of the typical American worker. You live by the sword, you die by the sword.

*****

(Total pay in my study usually differs from that shown in a proxy's Summary Compensation Table because it takes into account only the grant-date fair value of long-term incentive grants made in 2007, as opposed to including a partial charge for such grants in 2007 together with accruals for 2006 and earlier years. Compensation data for this study was obtained from Equilar Inc.)

(Graef Crystal is a columnist for Bloomberg News. The opinions expressed are his own.

To contact the writer of this column: Graef Crystal in Santa Rosa, California, at graefc@bloomberg.net.

Last Updated: March 26, 2008 00:01 EDT