there’s literally a strong causal relationship between marriage and having a shorter lifespan.
What causal relationship are you alluding to? As far as I can tell, the data you mention three comments above establishes a correlation between marriage and mortality, not causation. Moreover, that data also appears to show that the mechanism implicated in this correlation is complications during childbirth, which rules out marriage as the causal factor.
Hmm, so to be explicit, the claim I’m making is that marriage has a causal effect on mortality, mediated through complications in childbirth.
In Pearl’s do-calculus, this is
1. Marriage → greater rates of childbirth → Death.
I haven’t fully established this connection. The main way this argument falls is if it turns out marriage does not increase rates of childbirth. I assumed that marriage increases childbirth, but I admit to not looking into it.
I think when people are thinking about a strong causal relationship between marriage and mortality, they are mostly thinking of other mediating variables (weak claim, since most things are not childbirth). So:
2. Marriage → (collection of other mediating variables) → Death.
However, based on your and Liam’s comments, I’m starting to suspect that both of you mean causality in a much more direct sense. In that framework, perhaps the only relationship that will be considered “causal” should be:
3. Marriage → Death (no mediating variable).
If that is the case for your definition of causality, I agree that #3 is pretty unlikely. I also think it’s too strong since it probably rules out eg, smoking causing death (since you can’t use the mediating variable of lung cancer).
Your comment helped me understand this discussion better. It seems I was indeed assuming causality in the stronger sense, though I now see there wasn’t much justification for this assumption. As you point out, the stronger sense would fail to vindicate many relationships we generally take to be causal.
I still feel reluctant to assert that marriage causes death from the data you provided. Maybe it’s because I’m not sure what type of link exists between marriage and higher rates of childbirth. Though it seems clear that married people have more children, I’m not sure it’s correct to say that marriage causes people to have more children. People often get married because they want to have children. Even when this is not the initial motivation, it seems odd to say that marriage explains why these people have children. By contrast, the link between smoking and cancer seems much more tight.
I haven’t thought much about whether the causal attributions we make in social science tend to be more similar to “marriage causes higher rates of childbirth” or to “smoking causes higher rates of cancer”.
What causal relationship are you alluding to? As far as I can tell, the data you mention three comments above establishes a correlation between marriage and mortality, not causation. Moreover, that data also appears to show that the mechanism implicated in this correlation is complications during childbirth, which rules out marriage as the causal factor.
Hmm, so to be explicit, the claim I’m making is that marriage has a causal effect on mortality, mediated through complications in childbirth.
In Pearl’s do-calculus, this is
1. Marriage → greater rates of childbirth → Death.
I haven’t fully established this connection. The main way this argument falls is if it turns out marriage does not increase rates of childbirth. I assumed that marriage increases childbirth, but I admit to not looking into it.
I think when people are thinking about a strong causal relationship between marriage and mortality, they are mostly thinking of other mediating variables (weak claim, since most things are not childbirth). So:
2. Marriage → (collection of other mediating variables) → Death.
However, based on your and Liam’s comments, I’m starting to suspect that both of you mean causality in a much more direct sense. In that framework, perhaps the only relationship that will be considered “causal” should be:
3. Marriage → Death (no mediating variable).
If that is the case for your definition of causality, I agree that #3 is pretty unlikely. I also think it’s too strong since it probably rules out eg, smoking causing death (since you can’t use the mediating variable of lung cancer).
Your comment helped me understand this discussion better. It seems I was indeed assuming causality in the stronger sense, though I now see there wasn’t much justification for this assumption. As you point out, the stronger sense would fail to vindicate many relationships we generally take to be causal.
I still feel reluctant to assert that marriage causes death from the data you provided. Maybe it’s because I’m not sure what type of link exists between marriage and higher rates of childbirth. Though it seems clear that married people have more children, I’m not sure it’s correct to say that marriage causes people to have more children. People often get married because they want to have children. Even when this is not the initial motivation, it seems odd to say that marriage explains why these people have children. By contrast, the link between smoking and cancer seems much more tight.
I haven’t thought much about whether the causal attributions we make in social science tend to be more similar to “marriage causes higher rates of childbirth” or to “smoking causes higher rates of cancer”.
[Retracted]