Two Charts

Today I have two items with nothing in common except both struck me as interesting statistics.

1. Which adult American film actress was the highest paid movie star of the year four times?

.

.

.

.

Did you guess Doris Mary Ann Kappelhoff?

.

.

.

.

Better known as Doris Day, this actress was a top-ten movie earner in years 1951-1952 and from 1961-1966. She was the highest paid movie star in 1960, 62, 63 and 64. This puts her 6th place overall, tied with John Wayne and Shirley Temple who were also were the top star for four years.

Who's #1 on the list? Another surprise, at least to me. Tom Cruise has 7 years at the top of the pay scale. There is a four-way tie for 2nd place with 5 years as #1: Bing Crosby, Burt Reynolds, Clint Eastwood, & Tom Hanks

'I knew Doris Day before she became a virgin.'- actor Oscar Levant

2. The second statistic is an unusual graph at Axiis (go to the source to see the chart key). It shows the percentage market share of the major Web Browsers from 2002 to 2009. Normally one would show this as a bar graph or multiple pie charts. While this chart is clever I'm leaning towards not liking it. I find it very difficult to "see" the relative size of each browser and it's hard to "see" if a given percentage is 30% or 40%? The original image at Axiis is interactive; as you hover over a band, the percentage is shown.

Bottom Line

I love numbers. This weekend I met a man who does actuary science, the statistics of accidents and mortality. I explained I had a great admiration for statisticians but also a wariness that it was easy to mislead with stats. You can make convincing false arguments by manipulating the data.

For example, the Doris Day numbers cited above are a bit of fluke. She may have been #1 some year by just $1 over the next star or missed being #1 by the same amount. The yearly rankings don't tell us how much she won (or lost) by. It's a classic winner take all scenario much like the electoral college in US presidential elections. A President might have just 50.001% of the vote but take 49 of the 50 states (losing only the home state of his opponent).

How could star power be more accurately measured? Two ideas come to mind.

1. Measure their total earnings in inflation adjusted dollars. Possible problem? Have stars always earned the same percentage of a movie's gross? Like modern free-agent sports stars, did salaries at some point leap upwards (way above inflation) so that modern stars demand a higher share of ticket sales and hence skew the statistics?

2. Assign weights to the top ten rankings. Top Ten is absolute and immune to the salary scale spikes describe above. To avoid the arbitrariness of counting only #1, assign ten points for #1, nine points for #2, and so on down to 1 point for #10. Then add the points and see who comes out on top. This gives 2nd placers a chance to come out on top.