January 18, 2018

Family Structure, Race, and Crime

Some people, mostly conservatives, argue that a key factor in explaining racial crime disparities is racial differences in family structure. Specifically, they argue that the high rates of divorce and non-martial births among American Blacks explains why Blacks have higher crime rates than White Americans do. In this post we will look at few different things. First, we’ll look at the relationship between family structure and crime in general. Then, racial differences in family structure will be examined. And finally, we’ll look at what happens to the relationship between race and crime when you hold family structure constant. As will be seen, for several reasons, differences in family structure are probably not a significant cause of racial crime differences.

Family Structure’s Weak Relationship With Crime

The place to start this analysis is with the question “do people from broken homes, or children born out of wed lock, have higher than average crime rates?”. The answer is “yes, but barely”. Well and Rankin 1991 meta-analyzed 44 studies on the correlation between being from a broken home and juvenile delinquency. The mean N-weighted correlation was a mere .11. For violent delinquency the mean effect size across 6 studies was only .04. Similarly, Price and Kunz 2003 analyzed the relationship between family structure and juvenile delinquency across 72 studies and found a mean effect size of -.16. Moreover, this relationship decreased with age and was only -.10 among subject aged 16-19. More comprehensively,  Petrosine et al 2009 reviewed 5 previously meta-analyses which looked at the relationship between family structure and crime. The 5 correlations these meta-analyses produced between someone’s home breaking up and them later becoming a criminal were .07, .09, .09, .10, and .10. In other words, the effect is extremely weak and explains something like 1% of the population’s variance in criminality.

These studies are all comparing individuals. If we look at regional variation in crime the relationship between family structure and crime becomes a little stronger. For instance, Pratt and Cullen 2005 meta analyzed 137 studies on the relationship between crime, at a regional level, and family disruption, and found that 71.5% of studies found a significant effect with a mean effect size of .262. Similarly, Nivette 2011 meta-analyzed 10 studies which looked at how well divorce rates explained national variation in crime. The mean effect size was a statistically significant .277.

However, the regional relationship between crime over time, in the U.S. at least, has been extremely inconsistent. Back in the 1980s, proponents of the broken-homes-cause-crime theory would frequently point out that national rates of single motherhood began to rise around the same time violent crime began to spike in the 1960s. However, as reviewed in Luscombe 2012, the national relationship between single motherhood and crime fell apart in the 1990’s when violent crime tanked while non-marital birth rates continued to rise.

Since Mothers and Violent Crime 1960-2009

Thus, there doesn’t seem to be a strong case for family structure powerfully influencing crime even at the national level.

Genetic Confounding

More over, these effect sizes cannot be simply attributed to family structure causing crime. After all, parents who divorce, or have children out of wedlock, may be more aggressive, impulsive, etc., than average and may pass these traits onto their kids. The notion that marital instability is influenced by genetically impacted traits is supported by twin studies. For instance, McGue and Lykken 1992 found that having an identical twin who gets divorced increases your chances of getting divorced by 600% while having a fraternal twin or parent who gets divorced only increases your chances of being divorced by 200%. They estimated the overall heritability of divorce to be 52%. Indeed, it would be bizarre if divorce were not heritable since, as reviewed in Turkheimer 2000 Bouchard 2004 and Poldermane et al. 2015, virtually all human traits are heritable. From this point out, I am going to call the idea that genetically inherited traits explains the relationship between family structure and crime “the genetic-mediation hypothesis”.

Connors, Caspi, DeFries, and Plomin, 2000 attempted to test this possibility by comparing the effect of divorce on biological and adopted children. The reasoning behind this study was that adopted children don’t inherit genes from their legal guardians and, therefore, shouldn’t be made any worse off by them getting divorced on the genetic-mediation hypothesis. Their sample was rather limited. It included 188 adoptive families and 210 biological families. Child outcome measures consisted of variables associated with general adjustment and predictors of psycho-pathology. They were measured when the children were 12.

This first table reports on the measures of general psychological adjustment. The first 6 variables in this table measure how positively the child views themselves. The effect size, measured in standard deviations, was larger in biological families for 4 out of the 6 measures. However, the interaction between adoptive status and self-concept was not statistically significant. The next three measures are of “social competence” as reported by parents and the interviewer. Across all three measures, the effect size was larger in biological families. Two of these differences were statistically significant.  The last three variables measured achievement. In two of three cases, the biological effect size was larger. In biological families, the average effect size was .33. In adoptive families, the average effect size was .08. Thus, only 25% of the effect of divorce was still present after eliminating genetic con-founders.

Chart 1

The second chart shows the variables associated with psychopathology. As can be seen, parents in adoptive families reported fewer problems with their children while the opposite was true in biological families. However, the impact of divorce on substance abuse was much larger for adoptive families. The average effect size for biological families was .30. The average effect size for adoptive families was .29. So the effect sizes averaged across all five measures were very similar. However, the average difference for the 5 measures was .26, but in varying directions, so the previous statement is misleading. While genetic confounding did not account for most of the effect, the effects of divorce across the two family types were highly dissimilar.

chart 2

Despite some issues with statistical significance, no doubt resulting from the small sample size, I would say that this study lends support to the genetic-mediation hypothesis. The same evaluation could be made of Ryan et al 2013. Ryan et al. used a nationally representative sample of nearly 4,000 U.S. children and looked at how their behavior changed following a divorce.  They found that children in most income brackets and age groups did not exhibit a statistically significant increase in behavioral problems when their families shrunk from two parents to one.

Now that we’ve seen two studies which support the genetic mediation hypothesis, let’s look at two that don’t. Burt et al 2008 looked at how divorce among the parents of adoptive children predicted changes in behavioral problems. They found that adopted children from broken adoptive homes exhibited higher levels of delinquency, and that this was only true if the divorce happened after they had been adopted. That is, children adopted by parents who were divorced in the past but who did not divorce in the child’s life time did not exhibit high levels of behavioral problems. This not only runs counter to the genetic mediation hypothesis but also suggests that parents aren’t transmitting their divorce causing personality traits to their children via shared environment either.

Chart 3

Finally, I’d like to point you attention to Onofrio et al. 2005. This paper utilized what is called the children of twins study design. This involves comparing the children of two identical twins, one of which was divorced and the other of which was not, and seeing if the child of the divorced twin has more behavioral problems. You then do the same thing with children of fraternal twins. The idea is that the effects of divorce should be lower among the children of identical twins on the genetic mediation hypothesis because such children share more of their genes than the children of fraternal twins do. However, this is not what was found in Onofrio’s sample of 2,500 twin offspring. Thus, this study constitutes further evidence against the genetic mediation hypothesis.

On the whole, the evidence on this issue is mixed. Thus, the evidence based view is one of agnosticism about whether or not the weak association between family structure and crime can be interpreted causally.

Family Structure, Crime, and Race

Now that we’ve looked at the relationship between family structure and crime, let’s look at how family structure varies by race. First, let’s look at non marital births. Here, the differences are pretty simple. U.S. National Vital Statistics show that Blacks have the highest rates of non marital births, followed by Hispanics, followed by Whites, followed by Asians.

non marital births by race 2012

Moreover, Census Data shows that this is nothing new:

non martial births by race 1930 to 2008

The story with marriage is a little less straight forward. Today, data from the Bureau of Labor Statistics show that Blacks are less likely to ever get married than whites, and, because of this, slightly less likely to have ever been divorced. However, even when only looking at those who have been married, the difference in divorce rate is slight.

Marriage experience by race

Census data shows that this same basic pattern goes back several decades:

Marriage experience by race 1975-1990

However, this is not always been true. As reviewed in Ricketts 1989, prior to the 1950’s the Black marriage rate was actually higher than the White marriage rate.

Marraige rate by year by race

As can be seen below, data from the Department of Justice shows that Black over-representation in prisons was already present in the 1920’s and has grown since then. Given that divorce rates don’t differ much between the races, and the Black marriage rate was higher in the early 20th century when Blacks still had higher crime rates, if family structure does account for part of the gap then it must do so primarily via out of wed lock births, not divorce.

Chart 4

Given family structure’s weak relationship with crime, it would be surprising if it could account for much of the large gap in crime that exists between races. And, indeed, the relevant research suggests that it doesn’t.

Land, McCall, and Cohen 1990 collected data on the homicide rates of cities, standard metropolitan statistical areas (SMSAs), and states for the years 1960, 1970, and 1980. In each year they included all 50 states and every city and SMSA included in the census. They then looked at how well the following 11 variables predicted crime variation between these areas: population size, population density, percent black, percentage aged between 15 and 29, percent divorced, percent of kids without two parents, median family income, the poverty rate, income inequality, the unemployment rate, and whether or not the city/SMSA/State was in the south. All of these variables were entered into a single regression model, meaning that the estimated effect size for each variable held all other 10 variables constant. This analysis thus produced 9 total models explaining crime variation in cities, SMSAs, and states, across 3 decades. In all 9 models, race continued to predict homicide rates even while holding all of these variables constant. Moreover, in the majority of the models race was still a better predictor of violent crime than either the divorce rate or the percent of kids growing up without two parents.

Similarly, Kposowa, Breault, and Harrison 1995 analyzed crime variation across 3,076 U.S counties using a wide array of variables including the percent of the county that was divorced and the percent of the county that was black. One again, these variables are all being entered intro a single regression model and, so, the explanatory power of each variable which is reported reflect how well each variable predicted crime when every other variable was held constant. As can be seen below (standardized beta coefficients are under the “beta” column), the percent of the population that is black was a much stronger explanatory variable than the divorce rate for both property crime and violent crime:

chart 5

chart 6

Unfortunately, neither study tells us how large the association between race and crime was before adding all these other variables into the regression models. And, obviously, far more than family structure is being controlled for. However, the association between race and crime in these models is not small. Given this, we can say that a moderate to strong association between crime and race still exists even after controlling for a multiplicity of environmental explanations, including family structure.

Finally, it is worth mentioning that racial differences in family structure may, themselves, be partially genetic in origin. As reviewed else where on this site, the races differ with respect to sexual behavior such that Blacks tend to have more sex, with more people, produce more children, and be less monogamous. Given this, the fact that Blacks are less likely than Whites and Asians to raise their children under a two parents home is hardly surprising. This phenomena isn’t unique to the United States, either:

chart 7

Thus, the small impact on racial crime disparities that family structure may have might actually be one route by which genetic differences between the races lead to different crime rates.

In summary:

  1. Family structure is a weak predictor of crime
  2. Family structure in the U.S. has continued to deteriorate while violent crime has fallen
  3. It is possible that the weak relationship between crime and family structure which does exist is mediated by genetic factors
  4. The races differ in family structure. Blacks have always had more children out of wedlock and, since the 1950’s, have had lower marriage rates and higher divorce rates. However, difference in divorce rates are still rather small.
  5. Race is a better predictor of crime than family structure is and continues to predict crime when family structure is held constant.
  6. Given that Blacks are less monogamous than Whites all around the world, racial differences in family structure may be caused by genetics


Facebook Comments
  • Jared Huggins

    May I paraphrase Ann Coulter? “Of course crime rates are caused by single moms and illegitimacy, the (completely racially aggregated) data says so. Therefore the higher black crime rates are caused by illegitimacy (which they also happen to have a disproportionately high rate of).”

    Does she even realize her logic is completely circular?

  • Maxime Fansi

    Lol. And what about %of children in single parents households in Senegal, Ivory Coast, Mali, Cameroon, Angola etc… ?


      yes all those places have low crime rates lmao

      • Maxime Fansi

        All this place have low % of children in single parents households.


          no they dont. Do not kid yourself.

          • Maxime Fansi

            They do. You have to marry before having sex. And this is more true in west african countries which are muslim countries.

          • According to this paper, the biggest cause for single motherhood in sub-Saharan Africa is divorce or death “rather than as a consequence of premarital birth”.


            But I can’t find data on the specific countries you cite.

            Do you have sources?

  • David Lavery

    More sexism against fathers and men in general, feminism is a hate movement but my dad will always be special to me, I wish he was still alive.