Several responses:

Jonah,
Slow day here at work, so I ran your numbers.
Assuming almost half of 43 to equal 21 and about 13 to equal 15, the relationship is statistically not significant (by Fisher’s Exact Test). In fact, the p value is greater than 13% (while <5% is usually considered the point of significance). So, while it might be suggestive, you will have to wait until the sample size is considerably larger to see if the apparently higher prevalence of first borns among US Presidents is anything more than random chance.

And…

Jonah,
I don’t think that stat means much. This is my off-the-cuff, 2-minute so I don’t kill too much time version. By the way, I’m an actuary, which means I actually like this stuff.
Let’s start with simple distribution of family size similar to today’s (and I’m estimating here):
1/3 of families: 1 kid
1/3 of families: 2 kids
1/3 of families: 3 kids.
From the above distribution, if you select kids randomly, 50% are going to be first kids. There are 3 first kids out of 6 total kids; that’s where 50% comes from.
Now another question comes into play. When the study says firstborns, does that mean firstborn males only, or firstborn period?
If it means firstborn period, then we would actually expect on 25% firstborn males, so your study would be pretty significant.
If it means firstborn male, then we would actually expect the number of firstborns to be somewhat higher than 50% (using the above distribution) because a male could be born 2nd and still be the firstborn male.
Throughout history, family size has been larger, so these numbers would be reduced somewhat.
All-in-all, if it’s firstborn period, you’ve got a significant stat. If it’s firstborn male, I think you’ve got a “doesn’t mean much” stat.

And…

Please. While there might be a small statistical variation because families were bigger way back when, I would think that we have not reached a point where the numbers mean anything. I just rolled a dice 43 times. 14 sixes came up. That’s 33% of the time when statistically, the odds are that I’ll roll a six 16.6%. If I roll the dice a thousand times, maybe we’ll see a more even distribution.
I feel cheated. I want back the five minutes I put into this exercise. CNN should be ashamed; this falls under the WGAS banner anyway.

Me: One request: If you disagree with the number crunching on display in these emails, please ask yourself — before you email me about it — whether you so passionately disagree that it’s worth carrying on this conversation far outside the borders of actually being interesting.