Via Pandagon, I see that the San Francisco Chronicle has published an article about the silly claims of the likes of Arthur Brooks and Phillip Longman. You'd think that journalists could learn to be a little skeptical, but no, whatever these guys say gets reproduced without anyone bothering to check up on their sources. Prime example:
Take a randomly selected sample of 100 liberal adults and 100 conservative adults. According to an analysis of the 2004 General Social Survey -- a bible of data for social scientists -- the liberals would have had 147 kids, while the conservatives would have had 208. That's a fertility gap of 41 percent. Even adjusting for other variables like age and income, there is a gap of 19 percent.I meant to write another post about this last time the issue came up, but I ran out of steam after about 4 posts, so I just let it slide. But not this time.
You can go and check out the data for the General Social Survey yourself and see if Brooks' numbers can actually be found in the survey. They cannot. Below I post the results I get when I compare respondents' self-described political ideology to the number of children they have. The results are limited only to 2004 respondents:
-Column percent -Total pct -N of cases | POLVIEWS | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| 1 EXTREMELY LIBERAL | 2 LIBERAL | 3 SLIGHTLY LIBERAL | 4 MODERATE | 5 SLGHTLY CONSERVATIVE | 6 CONSERVATIVE | 7 EXTRMLY CONSERVATIVE | ROW TOTAL | ||
| CHILDS | 0: NONE | 46.4 1.4 18 | 42.6 3.9 50 | 36.2 4.5 58 | 27.8 10.5 136 | 31.1 5.1 66 | 17.5 3.0 38 | 18.7 .8 10 | 29.2 29.2 376 |
| 1: ONE | 8.2 .2 3 | 11.3 1.0 13 | 17.2 2.1 28 | 17.8 6.7 87 | 14.5 2.4 31 | 17.5 3.0 38 | 16.7 .7 9 | 16.2 16.2 209 | |
| 2: TWO | 22.2 .7 9 | 21.1 1.9 25 | 23.4 2.9 38 | 22.3 8.4 109 | 27.2 4.5 57 | 32.8 5.6 72 | 35.3 1.5 19 | 25.5 25.5 328 | |
| 3: THREE | 16.3 .5 6 | 9.4 .9 11 | 17.2 2.1 28 | 17.2 6.5 84 | 15.6 2.6 33 | 16.5 2.8 36 | 23.3 1.0 13 | 16.3 16.3 210 | |
| 4: FOUR | 5.9 .2 2 | 10.8 1.0 13 | 3.6 .4 6 | 8.4 3.2 41 | 7.2 1.2 15 | 9.8 1.7 21 | 6.0 .3 3 | 7.9 7.9 102 | |
| 5: FIVE | 1.0 .0 0 | 2.1 .2 2 | .6 .1 1 | 3.3 1.2 16 | 1.5 .2 3 | 2.4 .4 5 | .0 .0 0 | 2.2 2.2 28 | |
| 6: SIX | .0 .0 0 | 1.2 .1 1 | .5 .1 1 | 1.5 .6 7 | 1.2 .2 2 | 1.0 .2 2 | .0 .0 0 | 1.1 1.1 14 | |
| 7: SEVEN | .0 .0 0 | .0 .0 0 | .0 .0 0 | .8 .3 4 | 1.8 .3 4 | 2.0 .3 4 | .0 .0 0 | .9 .9 12 | |
| 8: EIGHT OR MORE | .0 .0 0 | 1.4 .1 2 | 1.2 .2 2 | 1.0 .4 5 | .0 .0 0 | .6 .1 1 | .0 .0 0 | .7 .7 10 | |
| COL TOTAL | 100.0 3.1 39 | 100.0 9.1 118 | 100.0 12.4 160 | 100.0 37.8 487 | 100.0 16.4 211 | 100.0 17.0 220 | 100.0 4.2 54 | 100.0 100.0 1,289 | |
| Means | 1.30 | 1.54 | 1.46 | 1.86 | 1.72 | 2.08 | 1.81 | 1.78 | |
| Std Devs | 1.41 | 1.74 | 1.50 | 1.69 | 1.59 | 1.59 | 1.17 | 1.62 | |
Now, there something you should notice right away. There's a little category called "moderate" that seems to have been ignored by these guys, in spite of the fact that it's overwhelmingly the largest single category. Indeed, the most salient result gleened just by looking at ideological self-indentification is that respondents overwhelmingly choose the middle categories, the dead-center being the most popular, and eschew the extremes. Yet Brooks simply divides everyone into "liberal" and "conservative", thus making American appear far more polarized than they actually are. Indeed, the whole article repeatedly commits this sin.
So having chucked out over 1/3rd of the data from the get-go, how does Brooks come up with his numbers? Who knows. They aren't to be found here. If we take all of the "liberal" respondents (all three categories, from extreme to slight) and do the math, we get an average of 1.47 children per liberal, which is what Brooks is reporting. However, if we do the same with the conservatives, we don't get 2.08. Instead we get 1.89. One possible explanation is that Brooks simply took those people who rated themselves a 6, which is "conservative", and used that number. If so, that's outrageously dishonest, because he doesn't do the same thing for liberals. Another possible explanation is that he used a different weighting method to come up with his result (I just used the default). But no matter which weighting method you use, you can't come up with 2.08 children per conservative. It's just not there.
Brooks claims that there is a 41% "fertility gap" between liberals and conservatives, but using the numbers that are actually in the GSS, the real gap is 29%. Given how precipitously these numbers fall when corrected for age and income, it's likely that there is no gap once you make these necessary corrections. And what if we did something radical, like actually included the more than 1/3rd of people who call themselves moderates, or decided to ignore the outliers? These things reduce (or possibly eliminate) the gap further.
As I've written about previously, even if people like Brooks weren't skewing the numbers, the differences between liberals and conservatives in terms of reproduction are miniscule. Back in the 1960s, fertility was so much higher than today that the "fertility gap" between someone then and now would be about 80%. Given just how little everyone reproduces these days, the differences we see are not enough to cause any significant demographic shift for many, many decades, by which time social and cultural changes will be far more important determinants of one's political views than whatever one's great, great grandfather may have believed. And perhaps most importantly, there is one very strong demographic trend occuring right now that conveniently never gets mentioned in these kinds of articles. It's called immigration. And it completely swamps out differences in native birth rates.
I would have hoped that mainstream journalists would have either gotten borded of this subject by now, or would have wised up and realized just how dishonest the people peddling it are. But sadly, that's not the case. This article tops most of the previous ones for stupidity with this bit stuck in the middle of the text:
[Podcast: Republicans are red-hot breeding machines]
I haven't yet listened to the podcast, but I don't need to to know how full of shit it's going to be. Let's look at the differences between Republicans and Democrats according the the GSS, going from 1996 to 2004:
| contain: -Column percent -Total pct -N of cases | PARTYID | |||||
|---|---|---|---|---|---|---|
| 0 STRONG DEMOCRAT | 1 NOT STR DEMOCRAT | 5 NOT STR REPUBLICAN | 6 STRONG REPUBLICAN | ROW TOTAL | ||
| CHILDS | 0: NONE | 25.0 5.9 504 | 28.0 8.9 760 | 29.1 7.7 660 | 21.0 3.8 323 | 26.3 26.3 2,247 |
| 1: ONE | 15.6 3.7 316 | 17.2 5.5 468 | 15.8 4.2 358 | 15.3 2.8 235 | 16.1 16.1 1,376 | |
| 2: TWO | 24.1 5.7 486 | 24.7 7.9 671 | 26.5 7.0 600 | 31.1 5.6 477 | 26.2 26.2 2,234 | |
| 3: THREE | 16.6 3.9 336 | 15.9 5.0 431 | 16.7 4.4 379 | 18.4 3.3 283 | 16.7 16.7 1,429 | |
| 4: FOUR | 10.3 2.4 208 | 7.1 2.3 192 | 7.0 1.9 160 | 8.1 1.5 125 | 8.0 8.0 685 | |
| 5: FIVE | 3.7 .9 75 | 3.1 1.0 85 | 2.6 .7 58 | 2.7 .5 42 | 3.1 3.1 261 | |
| 6: SIX | 2.2 .5 44 | 1.8 .6 50 | 1.0 .3 22 | 1.3 .2 20 | 1.6 1.6 136 | |
| 7: SEVEN | .9 .2 18 | .9 .3 23 | .8 .2 18 | 1.1 .2 16 | .9 .9 76 | |
| 8: EIGHT OR MORE | 1.6 .4 33 | 1.2 .4 33 | .5 .1 12 | .9 .2 14 | 1.1 1.1 92 | |
| COL TOTAL | 100.0 23.7 2,020 | 100.0 31.8 2,713 | 100.0 26.6 2,267 | 100.0 18.0 1,535 | 100.0 100.0 8,536 | |
| Means | 2.06 | 1.85 | 1.76 | 2.01 | 1.90 | |
| Std Devs | 1.80 | 1.72 | 1.58 | 1.60 | 1.68 | |
| Unweighted N | 2,043 | 2,695 | 2,250 | 1,537 | 8,525 | |
I've ignored the independents in the above table and just looked at those people who identify as either Republican or Democrat. Even without doing the math, it's obvious that Democrats have more children than Republicans (but not by much). Yet that doesn't stop Arthur Brooks from saying asinine things like this:
He [Brooks] reckons that unless something gives, Democratic politicians in the future may not have many babies to kiss.They'll have more babies than Republicans, you dumbass.
