Reverse Causality and Selection Bias

[Chinh Hoang-Duc]

Hi, I am doing a study to see how participating in commercial activities affects households' living standards. In the paper, I argue that the commercialisation decision is not random but must come from households' resources, thus, there is selection bias. I address that by using inverse probability weighting (IPW) controlling for confounders reflecting households' resources. Then two problem arises:

1. While including the resource variables in the IPW estimator, I find it difficult to differentiate them with ones I use to compute the living standard variables (income, housing, assets,...)
2. I also come across comments that what if these households started commercial activities many years ago causing their resources to differ from others, leading to reverse causation.

I am not sure if these two concerns are actually one. As my data is cross-sectional survey data, I can not do much about households' past activities and decisions.

So my questions are:

1. Is reverse causality the same as selection bias in this context? I am quite confused between these two as even if a household participated in this activity in previous years that caused changes in its resources, controlling for the resources in IPW at the current period can still eliminate the problem, right? If thinking in this way, I find that reverse causality is similar to selection bias.
2. If they are different, what should I do more to address reverse causality in my study? I have read some papers about it and IVs seem to be the best solution there is. Other methodological papers say that causal inference with reverse causality from cross-sectional data is a very contextual issue and it will vary from a dataset to another.

What do you think about these? I very much appreciate any help and discussion.