How much would the smaller urban numbers like the Beijing/Shanghai numbers be distorted by the fact that an official Chinese government study like the CFPS (run by the China Social Science Survey Center at Peking University) presumably only samples residents with hukous officially registered to the city, and omits the rest of the less elite population that actually live there?
It seems that they did sample people who did not have the right hukou, if my understanding of the system is correct. 20-40% or so of the population that claimed to live in Beijing/Shanghai said they had an agricultural hukou.
Hm, but you're getting substantially lower IQ scores than from your iodine dataset, aren't you? Your table says 109.7 for 'IQ' (I assume that's the "first principal component of the five tests") for 'Beijing', but then in your fertility chart Beijing is listed at ~112, which is higher, and I assume by iodine you mean the first entry for "北京" in https://gwern.net/doc/iodine/2006-li.pdf#page=3 at "114.1 依 14.6" (sampled way back in 2005) and averaging 109+114 gives your fertility chart entry.
And the difference between '107.8' for 'native' Beijingers and '101.6' for 'agricultural' Beijingers doesn't seem small to me: that's a third of a standard deviation from a single measure (which may itself be error-prone). At least 107 sounds more plausible than 114, anyway...
So maybe that's part of why the correlation is only r = 0.72: different sampling frames and changes in residency or migration over time. If you have selective migration where more intelligent residents of poorer provinces migrate to the elite cities (which of course they do) and there are samples which don't bend over backwards to sample from them, and further selection into who gets to be an official resident of elite cities, then you will exaggerate the differences by depleting the provinces while then not observing them in the elite cities.
Looking back at the sheet, the inconsistency is due to me copy-pasting the wrong column -- I have since updated the figures. The correlation is even lower (0.54) when using the correct column.
How much would the smaller urban numbers like the Beijing/Shanghai numbers be distorted by the fact that an official Chinese government study like the CFPS (run by the China Social Science Survey Center at Peking University) presumably only samples residents with hukous officially registered to the city, and omits the rest of the less elite population that actually live there?
It seems that they did sample people who did not have the right hukou, if my understanding of the system is correct. 20-40% or so of the population that claimed to live in Beijing/Shanghai said they had an agricultural hukou.
https://www.sebjenseb.net/p/answering-questions-on-regional-iqs
Hm, but you're getting substantially lower IQ scores than from your iodine dataset, aren't you? Your table says 109.7 for 'IQ' (I assume that's the "first principal component of the five tests") for 'Beijing', but then in your fertility chart Beijing is listed at ~112, which is higher, and I assume by iodine you mean the first entry for "北京" in https://gwern.net/doc/iodine/2006-li.pdf#page=3 at "114.1 依 14.6" (sampled way back in 2005) and averaging 109+114 gives your fertility chart entry.
And the difference between '107.8' for 'native' Beijingers and '101.6' for 'agricultural' Beijingers doesn't seem small to me: that's a third of a standard deviation from a single measure (which may itself be error-prone). At least 107 sounds more plausible than 114, anyway...
So maybe that's part of why the correlation is only r = 0.72: different sampling frames and changes in residency or migration over time. If you have selective migration where more intelligent residents of poorer provinces migrate to the elite cities (which of course they do) and there are samples which don't bend over backwards to sample from them, and further selection into who gets to be an official resident of elite cities, then you will exaggerate the differences by depleting the provinces while then not observing them in the elite cities.
Looking back at the sheet, the inconsistency is due to me copy-pasting the wrong column -- I have since updated the figures. The correlation is even lower (0.54) when using the correct column.
These were the averages from the Iodine study -- I double checked with Richard Lynn's publication on China (https://www.sciencedirect.com/science/article/abs/pii/S0160289616301805). All were eventually subtracted by 3.4 as the average score on the raven's was 103.4.
Beijing 114.1
Tianjin 105.3
Hebei 105.4
Shanxi 108
Inner mongolia 105.1
Liaoning 107.5
Jilin 107
Heilongjiang 101.4
Shanghai 115.3
Jiangsu 109
Zhejiang 115.8
Anhui 98.2
Fujian 107.1
Jiangxi 98.9
Shandong 107.9
Henan 95.4
Hubei 105.3
Hunan 103.8
Guangdong 101.1
Guangxi Zhuang Autonomous Region 98.3
Hainan 90.7
Chongqing 106.3
Sichuan 105.4
Guizhou 92.8
Yunnan 96.8
Tibet 77.3
Shaanxi 104.7
Gansu 96.9
Qinghai 92.8
Ningxia Hui Autonomous 93.4
Xinjiang Uygur Autonomous Region 98.2
I could try looking into that.
Nice variable names