A few years back, the SEC passed a rule forcing publicly traded companies to disclose the ratio between CEO compensation and median worker pay. Many observers thought the rule would go away with President Trump’s election, but they were wrong—companies are now disclosing their pay ratios before a June 30 deadline. And surprise! The results so far don’t look good for many companies in the S&P 500.

Our visualization highlights the 20 biggest pay gaps in America according to data collected by the AFL-CIO. We started by placing a photo of the CEO and the name of their company directly above their total compensation. We then added the median worker salary, illustrating the different pay ratios with a series of corresponding circles. The bigger the circle, the bigger the ratio. This is a straightforward way of representing the data, and it reveals several key insights into executive compensation in corporate America.

Let’s start by getting clear on how to read the chart. For example, Mattel has the biggest circle on the visualization. For every $1 an average worker makes, the CEO takes home $4,987. Several companies pay their CEOs well over 1,000 times as much as the average worker. However, overall CEO compensation does not exactly follow the ratio. In fact, Richard Fahn, the CEO of Royal Caribbean Cruises, has the lowest pay ratio on our visualization at $728. But overall he gets paid more than 9 other chief executives on the list. That’s because Royal Caribbean compensates its employees better than other companies. In fact, the best company for average workers is Wynn Resorts, where the average person takes home $44,437. That’s pretty good all things considered, but not nearly as good as the $34.5M that Stephen Wynn takes home.

We should also point out how there are several important underlying assumptions behind the data. In fact, on page 17 of its actual rule, the SEC explicitly refuses to define the term “employee,” only stating that it includes anyone employed on the last day of the previous fiscal year whether in a “full-time, part-time, seasonal or temporary” capacity. Each company is also “permitted but not required” to annualize each employee’s compensation, which further complicates the data. CEO pay also include some fuzzy numbers, like the expected future value of unvested stock or the value of non-monetary forms of compensation, like a CEO’s security detail. Both factors tend to inflate the ratios, especially in the retail sector (where there is high seasonal employment).

It’s worth pointing out that even if these numbers are inflated, they still represent an enormous disparity in compensation. If Mattel’s CEO was paid only 3,000 times more than the average worker as opposed to 4,000 times more, would it be any less astonishing?