Fair500 → Analysis → Reading the table
The Summary Compensation Table is where executive pay actually lives. It is also laid out in a way that reliably produces two specific mistakes, both of which we corrected during verification.
If you want to know what a chief executive was paid, there is exactly one authoritative place to look: the Summary Compensation Table in the company's DEF 14A proxy statement, filed annually with the SEC and freely available on EDGAR. Everything reported in the press traces back to this table.
It is a wide grid, usually rotated sideways on the page, showing several named executives across three fiscal years. This is how to read it, and how to avoid the errors it invites.
The table's columns are prescribed by SEC rules, so they are consistent across filers even when the layout is not.
| Column | What it contains |
|---|---|
| Salary | Base cash salary actually paid in the fiscal year. Usually the smallest component for a large-company CEO. |
| Bonus | Discretionary cash bonuses only. Most performance bonuses appear under Non-Equity Incentive instead, so this column is frequently zero. |
| Stock Awards | Grant-date fair value of share awards granted this year. Usually the largest number in the table. |
| Option Awards | Grant-date fair value of stock options granted this year. |
| Non-Equity Incentive Plan | Cash paid out under a formal performance plan, which for most companies is the annual bonus. |
| Change in Pension Value | Increase in actuarial value of pension and deferred compensation. Can be large and can move on interest rates rather than any decision. |
| All Other Compensation | Perquisites, insurance, retirement contributions, personal aircraft use, security. Detailed in a footnote. |
| Total | The exact sum of the preceding columns. This is the figure used in the pay ratio. |
Stock and option awards are recorded at grant-date fair value, an estimate made on the day of the grant of what the award is worth. It is not what the executive received in cash, and it may never become that.
Two consequences follow. An executive whose company collapses can show enormous disclosed compensation for awards that vest worthless. An executive whose company triples can realise far more than the disclosed figure. And because the entire value of a multi-year grant is booked in the year it is granted, a single year's total can be wildly unrepresentative, which is why Fair500 averages CEO pay over three years.
If you want realised pay instead, look at the Option Exercises and Stock Vested table, which appears later in the same proxy. It shows what actually vested and was exercised during the year. It answers a different and often more interesting question than the Summary Compensation Table does.
There is a strong check available, and it is the single most useful technique we found: in a valid Summary Compensation Table, the Total column is the exact sum of the component columns. Add up salary, bonus, stock, options, non-equity incentive, pension change and other compensation. If the result does not equal the stated total to the dollar, you have misread a row, picking up a number from an adjacent year, skipped a column, or landed on the wrong executive.
We used this as the primary validation when compiling data for 494 companies, and roughly 197 passed it outright. It is a demanding test, and one that a misidentified row is very unlikely to pass by chance.
One warning from experience. Our first implementation accepted a match within 0.5% of the total, on the theory that rounding in the published figures might cause small discrepancies. On a $28 million package, 0.5% is a $140,000 tolerance, wide enough to accept rows that were simply wrong. Tightening the tolerance to one dollar changed a number of our figures. A tolerance expressed as a percentage of a large number is not a check.
This is the error we most want to warn people about, because it is easy to make, produces a plausible-looking number, and is invisible unless you check.
Executives are typically listed in order of seniority or compensation, and a great many companies list an executive chairman or the chief financial officer first. Taking the top row therefore attributes the wrong person's pay to the chief executive. In our data this affected Intel, where the top row is the finance chief rather than the CEO, and the same pattern appeared at Diamondback Energy, Tyson Foods and Rollins.
The fix is to read the title column rather than the position on the page, and to be alert to titles like "Executive Chairman", "Founder and Chairman" and "President and CFO" that can look authoritative at a glance.
When a chief executive changes mid-year, both appear, each with partial-year compensation, and the ratio the company discloses may relate to either. The rule of thumb we settled on: use the person who served as chief executive for the most recent reported fiscal year, which is frequently not the person holding the job on the day you read the filing. Target's most recent reported year belongs to Brian Cornell even though a successor has since taken over; the same applied at Prologis and Camden Property Trust.
Where a company has genuine co-chief-executives, as KKR and Netflix do, we average the two. That matches KKR's own disclosure practice. Summing them would imply a single person's pay and roughly double the true figure.
The ratio is usually a short standalone paragraph headed "CEO Pay Ratio" or "Pay Ratio Disclosure", placed after the compensation tables. It states the median employee's annual total compensation, the CEO's, and the ratio, followed by a description of the methodology used to identify the median employee. That methodology paragraph is worth reading. It tells you whether the company sampled, which compensation measure it ranked on, what date it used, and whether it excluded any foreign employees.
Do not assume the ratio is in the proxy at all. Three companies in the S&P 500 defeated our first pass for structural reasons:
Newly public companies are a fourth case: a company in its first year after an IPO or spinoff is exempt from the disclosure entirely, which is why five S&P 500 constituents are absent from this site.
Everything above is free. Go to EDGAR full-text search, enter the company name or ticker, and filter to form type DEF 14A for the proxy or 10-K for the annual report. The most recent proxy contains three years of compensation history, so two filings give you six years.
One small technical note for anyone parsing these documents programmatically: proxy HTML frequently contains zero-width space characters (U+200B) inside numbers and headings, which will silently defeat string matching. Strip them before anything else. It cost us an afternoon.