If your parents are on welfare

your chances of ending up on welfare yourself are higher. So what? The link is only interesting if it is causal -- and unobservables can very well be expected to drive the correlation across generations. For example, people who struggle with education will probably have children who also don't do well in school; their chances of requiring hand-outs must be higher.

How to solve the problem? Norwegian data to the rescue! Dahl et al. in new NBER working paper use random assignment to judges in Norway - with some of them more likely to uphold/deny access to disability insurance -- to identify the nature of the link. The result - it's causal, it's big, and it's depressing:

Strong intergenerational correlations in various types of welfare use have fueled a long standing debate over whether welfare dependency in one generation causes welfare dependency in the next generation. Some claim a culture has developed in which welfare use reinforces itself through the family, because parents on welfare provide information about the program to their children, reduce the stigma of participation, or invest differentially in child development. Others argue the determinants of poverty or poor health are correlated across generations, so that children's welfare participation is associated with, but not caused by, parental welfare use. However, there is little empirical evidence to sort out these claims. In this paper, we investigate the existence and importance of family welfare cultures in the context of Norway's disability insurance (DI) system. To overcome the challenge of correlated unobservables across generations, we take advantage of random assignment of judges to DI applicants whose cases are initially denied. Some appeal judges are systematically more lenient, which leads to random variation in the probability a parent will be allowed DI. Using this exogenous variation, we find strong evidence that welfare use in one generation causes welfare use in the next generation: when a parent is allowed DI, their adult child's participation over the next five years increases by 6 percentage points. This effect grows over time, rising to 12 percentage points after ten years. Using our estimates, we simulate the total reduction in DI participation from a policy which makes the screening process more stringent; the intergenerational link amplifies the direct effect on parents at the margin of program entry, leading to long-run participation rates and program costs which are substantially lower than would otherwise be expected. The detailed nature of our data allows us to explore the mechanisms behind the causal intergenerational relationship; we find suggestive evidence against stigma and parental investments and in favor of children learning from a parent's experience with the DI program.