Have I mentioned how much I love Scandinavian national registries? They enable some amazing research. Here is an example: "Childhood family income, adolescent violent criminality and substance misuse: quasi-experimental total population study", Sariaslan et al 2014 https://pdf.yt/d/oUgs1U5suhiilEPi / https://dl.dropboxusercontent.com/u/182368464/2014-sariaslan.pdf (media: http://www.economist.com/news/science-and-technology/21613303-disturbing-study-link-between-incomes-and-criminal-behaviour-have-and )

This investigates poverty/SES and crime. As I understand the design: one of the problems with investigating causes of crime and poverty is that it's hard to measure or control for heredity and families (which do matter although not as much as people think). You could use an adoption study of twins or something but there's never enough to go around. But you can (more or less) control for genes and family environment if you take the same family and make it richer/poorer, and see what happens to kids born in different conditions (if kid #1 is born while they're poor and then you randomly hand them a fortune so they are rich when kid #2 is born, then all the kid \#1s should do much worse than kid \#2s if poverty itself causes crime & other problems). This way, the family is the same (by definition) and the kids are pretty similar geneticly (same parents, even if they're not twins). That experiment is hard to do (where does one get that much money to hand out?), so you can fall back to observing families becoming richer/poorer on their own and hope the quasi-natural-experiment works. But that's hard to do too since incomes are fairly stable so you'd need to observe thousands upon thousands upon thousands of families to see some families go up/down and then observe whether they differ on fairly rare events like criminality.

Enter... the Scandinavian national registries. They have data on thousands upon thousands upon thousands of families.

The results are striking, especially for an environmentalist perspective; excerpts:

"_Background_: Low socioeconomic status in childhood is a well-known predictor of subsequent criminal and substance misuse behaviours but the causal mechanisms are questioned.
Aims: To investigate whether childhood family income predicts subsequent violent criminality and substance misuse and whether the associations are in turn explained by unobserved familial risk factors.
Method: Nationwide Swedish quasi-experimental, family-based study following cohorts born 1989–1993 (n_total = 526 167, n_cousins = 262 267, n_siblings = 216 424) between the ages of 15 and 21 years.
Results: Children of parents in the lowest income quintile experienced a seven-fold increased hazard rate (HR) of being convicted of violent criminality compared with peers in the highest quintile (HR = 6.78, 95% CI 6.23–7.38). This association was entirely accounted for by unobserved familial risk factors (HR = 0.95, 95% CI 0.44–2.03). Similar pattern of effects was found for substance misuse.
Conclusions: There were no associations between childhood family income and subsequent violent criminality and substance misuse once we had adjusted for unobserved familial risk factors.

Recently, a Norwegian total population study found that children of parents in the lowest income decile were twice as likely to be convicted of a violent or drug crime compared with their peers in the fifth decile. 3 Similarly, a number of longitudinal USA studies have linked low income levels with substance use disorders. 4,5 Nevertheless, these findings could potentially result from inadequate adjustment of familial risk factors. 6 Behavioural genetic investigations have found that the liabilities for both violent offending and substance misuse are substantially influenced by shared genetic and, to a lesser extent, family environmental factors. 7,8 Consequently, it has been proposed that quasi-experimental, genetically informative research designs that explicitly take such factors into account could be integral in elucidating the causal mechanisms further. 9

- 6: Sariaslan A, Langstrom N, D’Onofrio B, Hallqvist J, Franck J, Lichtenstein. "The impact of neighbourhood deprivation on adolescent violent criminality and substance misuse: a longitudinal, quasi-experimental study of the total Swedish population" http://www.researchgate.net/publication/256985352_The_impact_of_neighbourhood_deprivation_on_adolescent_violent_criminality_and_substance_misuse_A_longitudinal_quasi-experimental_study_of_the_total_Swedish_population/file/72e7e51f69f99ad646.pdf . Int J Epidemiol 2013; 42: 1057–66.
- 7: Frisell T, Lichtenstein P, Langstrom N. "Violent crime runs in families: a total population study of 12.5 million individuals" http://www.sakkyndig.com/psykologi/artvit/frisell2010.pdf . Psychol Med 2011; 41: 97–105.
- 8: Kendler KS, Sundquist K, Ohlsson H, Palmér K, Maes H, Winkleby MA, et al. "Genetic and familial environmental influences on the risk for drug abuse: a national Swedish adoption study" http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3556483/ . Arch Gen Psychiatry 2012; 69: 690–7.
- 9: D’Onofrio BM, Lahey BB, Turkheimer E, Lichtenstein P. "Critical need for family-based, quasi-experimental designs in integrating genetic and social science research" http://people.virginia.edu/~ent3c/papers2/d%27onofrioAJPH.pdf . Am J Public Health 2013; 103: S46–55.

We linked data from nine Swedish, longitudinal, total-population registers maintained by governmental agencies. The linkage was possible through the unique 10-digit civic registration number assigned to all Swedish citizens at birth and to immigrants upon arrival to the country...The following nine registers were used: (a) the Total Population Register (TPR) contained basic information (for example, gender and date of birth) for all individuals registered as inhabitants of Sweden; (b) the Multi-Generation Register supplied data that linked index individuals found in the TPR to their biological parents, thus enabling us to connect siblings; (c) the Medical Birth Register included pregnancy data with close to full coverage (499%) of all births in Sweden since 1973; 14 (d) the Education Register contained information on highest level of completed formal education; (e) the Cause of Death Register provided data on principal and contributing causes of death since 1958; (f) the Migration Register supplied data on dates for migration into or out of Sweden; (g) the Integrated Database for Labour Market Research (LISA) provided annual information on family disposable income and welfare recipiency since 1990 on all individuals 16 years of age and older who were registered in Sweden as of December 31 for each year; (h) the National Patient Register provided data on psychiatric in-patient care since 1973 (ICD-8, -9 and -10) 15–17 and out-patient care since 2001 (ICD-10); and (i) the National Crime Register supplied detailed information on all criminal convictions in lower general court in Sweden since 1973. Plea bargaining is not allowed and conviction data include all individuals who received custodial or non-custodial sentences; also those cases where the prosecutor decided to caution or fine. Only individuals age 15 or older are legally responsible in Sweden; hence, we were not able to study criminal offending prior to age 15. A total of 594 127 children were born in Sweden between 1989 and 1993 and registered in the Medical Birth Registry. We chose to exclude children from multiple births (n = 14 670), those who had serious malformations at birth (n = 20 905) or who could not be linked to their biological parents (n = 3 956). Furthermore, we excluded data for children who had either died (n = 2 525) or emigrated from Sweden before they reached 15 years of age (n = 18 301). Last, we removed individuals with missing data on parental labour market exposures (n = 7603).

We calculated mean disposable family income (net sum of wage earnings, welfare and retirement benefits, etc.) of both biological parents for each offspring and year between 1990 and 2008. Income measures were inflation-adjusted to 1990 values according to the consumer price index provided by Statistics Sweden (http://www.scb.se/en_/). Econometric researchers have long recognised that single annual income exposure measures generally suffer from substantial measurement error because of their inability to accurately depict long-term SES, often leading to attenuation bias. 18,19 Therefore, annual variables were used to calculate the mean parental income throughout each offspring’s childhood (ages 1 through 15).
Early critics challenged the linearity assumption used by studies adopting continuous income measures by contending that criminality is largely confined to the lowest social strata, often referred to as ‘the underclass’ or ‘the poor’, with little to no difference being found between the strata in the mid to upper ranges of the income distribution. 20 Others have argued that the cause of the spurious correlations are because of separate mechanisms promoting deviant behaviours on both ends of the income distribution resulting in weak mean predictions. 1 We decided, therefore, to test potential non-linear effects by categorising our income measure in quintiles.

Violent crime was defined as a conviction for homicide, assault, robbery, threats and violence against an officer, gross violation of a person’s/woman’s integrity, unlawful threats, unlawful coercion, kidnapping, illegal confinement, arson, intimidation, or sexual offences (rape, indecent assault, indecent exposure or child molestation, but excluding prostitution, hiring of prostitutes or possession of child pornography). 21
In line with previous studies using Swedish total population data, 8,22 we used an omnibus measure of substance misuse consisting of convictions of any drug-related crimes (defined as crimes against the Narcotic Drugs Act (SFS 1968:64) or driving under the influence of alcohol and/or illicit substances) or having been diagnosed with an alcohol- or drug-misuse-related disease in in-patient or out-patient settings

To assess the effects also of unobserved genetic and environmental factors, we fitted stratified Cox regression models to cousin (n = 262 267) and sibling (n = 216 424) samples with extended or nuclear family as stratum, respectively. The stratified models allow for the estimation of heterogeneous baseline hazard rates across families and thus capture unobserved familial factors. 23 This also implies that exposure comparisons are made within families. 24 Model III was fitted to the cousin sample and adjusted for observed confounders and unobserved within extended-family factors. Model IV was fitted on the sibling sample and accounted for unobserved nuclear family factors and for gender, birth year and birth order.

Table 2 presents results from multivariable Cox regression models; children of parents in the lowest income quintile had an almost seven-fold increased hazard of being convicted of violent crime (crude HR = 6.78, 95% CI 6.23–7.38) and a two-fold increase of substance misuse (HR = 2.45, 95% CI: 2.32–2.58) in adolescence compared with peers whose parents were in the fifth quintile (Model I).
When we made adjustments for observed family-wide risk factors (Model II), the effects of childhood family income on violent criminal convictions were significantly attenuated but remained strong (HR = 3.93, 95% CI 3.59–4.30). Controlling for family-wide risk factors also affected the association with substance misuse (HR = 1.98, 95% CI 1.86–2.10). Model III expanded on Model II by also accounting for unobserved familial risk factors within extended families through cousin comparisons. This adjustment reduced the hazard ratios by 50% and 25% for adolescent violent crime and substance misuse, respectively. Finally, we studied the effects of unobserved familial risk factors within nuclear families using sibling comparisons (Model IV). The associations between childhood family income and the outcomes disappeared completely; hazard ratios were 0.95 (95% CI 0.44–2.03) for violent crime and 1.11 (95% CI 0.62–1.98) for substance misuse, respectively. This suggested that unobserved familial factors fully accounted for the increased hazard ratios found in previous models.

Using traditional epidemiological methods, we found that low income in one’s family of origin was indeed associated with higher risk of violent offending and substance misuse during adolescence. However, the excess risks became marginal or disappeared completely when we gradually adjusted for familial risk factors of these associations by studying within-extended family and within nuclear-family estimates (with cousin and sibling controls, respectively). This held true when childhood SES was defined either as parental disposable income or welfare recipiency throughout child ages 1–15 years. Sensitivity analyses proved the results were robust across gender, ethnicity and age periods and were not influenced by limited within-family variability in the exposure variables.

Third, the sibling-comparison design makes several important assumptions and requires a large sample size. 9,37,38 In principle, only sibling pairs discordant on both exposure and outcome contribute to the analyses. We identified 116 875 siblings in 56 551 families who were discordant for childhood family income (measured in deciles). Among these discordant siblings, 3195 were further discordant for violent criminal convictions and 5507 for substance misuse. Although they might seem small, these sample sizes are still larger than in most of the previous studies. Moreover, the sibling- comparison design assumes that the results of discordant siblings are generalisable to the total population. We found no income differences when comparing the discordant siblings to the total population; t(526 165) = 1.25, P = 0.21. Thus, our findings do not seem to follow from poor statistical power, neither does it seem that results from discordant siblings are not generalisable."