Key word is “only.” Meaning no variance. There is some variance.
A predictor has variance.
We've used the word "correlated" here, but that's not the right metric. Success might be correlated with a nuclear family when you just look at people and outcomes, but that doesn't mean that the nuclear family *caused* that outcome, any more than Asian parents. And since this is a "predictor" and not a "certainty" - it's not really that it *caused* it but more so *influenced* it.
The nuclear family hypothesis looks good because of the correlation - but it's not the nuclear family itself - it's that nuclear families are a predictor of financial stability - and the financial stability is the real causal predictor. If you take poor nuclear families, you'll get worse outcomes. If you take rich non-nuclear families, you'll get better outcomes.
I mean sure, there are other predictors. Downs Syndrome is a predictor that your outcome will be lower on the scale. But money swamps it all.