ssa_national set is not gender balanced

[bmschmidt]

method="ssa" could probably use some adjustment to compensate for something I noticed in the Social Security dataset: before about 1918, it's two-thirds women.

I assume this has to do with who was eligible for benefits when the program was created in the late 30s: either the men born around 1900 are dead, or more likely they're not eligible for survivor benefits for spouses or something.

All the ratios for years around 1900 from this method are distorting the female % of the name: for example, in 1901 merle has 91 women and 52 men counted, but since 69% of the sample is female that year, that male number should be 2.2x higher: the right prediction that it's male, not female.

An illustrative plot:

gender::ssa_national %>% group_by(year) %>% summarize(ratio = sum(female)/sum(female+male)) %>% ggplot(aes(x=year,y=ratio)) + geom_line() + labs(title="Percentage of the set that is women")

[lmullen]

Thanks for pointing this out, @bmschmidt. I'll have to take it into account.

[lmullen]

First stab at a solution with justification for the reasoning: http://rpubs.com/lmullen/gender-imbalance-ssa

[lmullen]

Fixed on develop.