Youth suicide

Explaining The Rise In Youth Suicide

The following is a summary of “Explaining the Rise in Youth Suicide”, by David M. Cutler, Edward Glaeser, Karen Norberg.

This article is the best we have found for comparing suicide theories to what the data actually says. It is essential reading because the US data is richer than in many other countries, permitting more insights that may also have applicability outside of the USA.

A key characteristic of the US data is that suicide rates amongst US youths aged 15-24 has tripled in the past half-century, even as rates for adults and the elderly have declined. At 13/100,000 suicide is the 3rd leading cause of death, after accidents (which may include some suicides) and homicide. And for every youth suicide completion, there are nearly 400 suicide attempts. Other observations on US data are:

  • Suicide and homicide are positively correlated.
  • Girls attempt suicide more than boys, but boys complete it more than girls.
  • Suicide clearly decreases with age after adolescence (one reason why The Anika Foundation has chosen to focus on adolescent depression).
  • Blacks attempt and complete fewer suicides than whites.
  • Rural states have higher suicide rates.
  • Completed suicides are overwhelmingly accomplished with guns in the US. etc.

Some of the theories of youth suicide the authors explore are:

  1. ‘Rational’ youth suicide: for example, because of chronic depression. Youths make the judgement that their suffering is such that the unhappiness of life is valued much less than death.
  2. Strategic behaviour on the part of youths. Youths have little direct economic or familial power. In such situations, self-injury can be used by youths to signal distress, or to encourage a response by others. Attempts to punish altruistic parents or other adults are a key part of this.
  3. Instrumentality (e.g. the ready availability of guns etc), combined with impulsive behaviour.
  4. Contagion theories based on imitative behaviour (as in Durkheim’s early views).

The authors look at these theories, each of which have various sub hypotheses and subtleties, from three quite different data sets: first, that on attempted suicides, using survey information from the AddHealth data; second, since surveys are not possible for completed suicides, they use national versus county data variations in suicides and related explanatory data; and third, they look at changes over time using detailed state data.

But Remember;

First, the overriding data-versus-theory issue that should always be kept in mind is this: whatever your theory of youth suicide is, it must be consistent with explaining why it’s rate of incidence is rising so strongly over time. A puzzle indeed.

Second, all analysis will be plagued by problems of endogeneity: what is really causing what? And are there missing variables and hypotheses that need to be thought about?

 

The Suicide Attempt Data

With regard to suicide attempts, using the AddHealth data permits medically screened as well as total suicide attempts to be compared to matched data concerning:

  • demographics (female, black, Asian, native American, urban);
  • age characteristics (from 12 to 18 years);
  • employment and income (including mother’s participation in the work force);
  • family structure (mother’s presence, father’s presence, never knew father, knew father but he is no longer there, stepfather);
  • interaction with parents (relationship with mother, father and non-resident father);
  • sexual activity (intercourse, rape);
  • delinquency (including violent behaviour, or violence against the person);
  • drug and alcohol problems; participation (in clubs, sports, TV watching hours);
  • contagion (eg friend attempted suicide, friend completed suicide, relative attempted, relative completed);
  • and whether the person suffered from depression (score used).

The regression results for suicide attempts are similar for the medically screened and total samples. The results for total attempts, where all of the explanatory variables are included together, show the following relative importance (statistical significance at the 96% level).

1st
(by far the most significant), the depression score
2nd
drug use
3rd
friend attempted suicide
4th
raped
5th
delinquency
6th
violent behaviour
7th
relative attempted suicide
8th
friend died by suicide
9th
alcohol problems
10th
a poor relationship with a non-resident father
11th
being 14 years of age
12th
being female

(The other variables mentioned earlier, but not included in these 12, were only marginally significant or not significant at all.)

If the key depression score is used as the dependent variable, then of the variables studied, the highest correlated factors (in order of importance) were: the delinquency score; being female; being Asian; having a friend who suicided; having been raped; having been hurt in violence; a poor relationship with the mother; a poor relationship with a resident father; alcohol problems; the family being on welfare; drug problems; a relative who attempted suicide; low family income; being 13 years of age; low participation in clubs/societies; father not home in the evening; being native American; being black; mother not home in evening; stepfather in the home (surprisingly negatively correlated with depression-good news for step dads); and being an urban dweller.

 

Suicide Completion Data Across Countries

The authors note that the motivations for suicide completions are very different to attempted suicide, (most people who complete actually do intend to die). However, the above interview-based data is not available for completions. So the authors use detailed county data across the USA, looking for explanations of variations in county suicide completions versus the national rate.

The explanatory variables are: urbanicity (large urban, small urban, farm, population greater than 1m); demographics (black, native American, Asian); economic (median income); social characteristics (% divorced females, female employment rate, share of step children, share of female head of families); and gun ownership (shared owning generally, and shared gun ownership for hunting purposes, i.e. rural/farm).

The most important explanatory variable by far is the increased share of youths living in homes with a divorced parent.  Next came being native American. Then there was access to shared guns for hunting (which, when added to the regression, knocked out the otherwise powerful factor of living on a farm, low income, and guns shared more generally). If the rise in the divorce rate over time were simply superimposed on their correlations, then mechanically this would explain 2/3 of the rise in youth suicide over time.

 

Changes in State Suicide Completions Over Time

But when it came to running the regressions for data that actually includes changes over time, based on state suicide data and statistics, then the divorce rate actually dropped out. The clearest explanatory variables were: the rise in female participation in the labour market, and the change in family median income (higher is better). Female participation rising and low family income is probably highly correlated with divorce.

This aspect of the research opens the door for economic hardship and more general social stressors to enter. Divorce, rising female participation, and suicide might be related to a third set of factors, such as increasing economic hardship (e.g. through widening income disparity and competition for jobs that pay), that cause social stress to rise, put pressure on marriages, necessitate higher female participation, etc.

 

Our Comments On The Suicide Theories That Follow

First, we agree with the authors that there is some support for the theory that many suicide attempts by youths (as very distinct from completions) can be viewed as a strategic action to resolve conflicts within oneself (delinquency, drug use, sexual activity, rape, alcohol, victimisation), or to punish others (the non-resident father variable, and the poor relationship with mothers, via the depression channel of influence).  Being female also seems to be highly correlated with the strategic attempts at suicide, as opposed to completions (where males dominate).

Second, we agree with the authors that there is clear evidence for contagion effects. Youths who have a friend or family member who attempts or commits suicide are more likely to attempt or commit suicide themselves. There is individual-level evidence of contagion in the US AddHealth data, and clear statistical evidence of non-random clustering in the country-level vital statistics (e.g. across counties including urban and remote regions). Contagion may involve the direct influence of one teen’s suicidal behaviour on another, or it may involve more indirect social and cultural processes. But in either case these “neighbourhood effect” may multiply the effects of events (a negative) and government policies (potentially positively).

Third, the instrumentality/impulsiveness theory is not that well supported in this study. Allowing for guns used for hunting knocked out general gun ownership. Guns have always been widely available for hunting in rural areas, but it is gun ownership in urban areas that has grown over time during the Post-War period. (Remember, suicide theories must explain why there has been a rise in the youth suicide rate over time.) In any case, there are many countries with less gun ownership that have higher suicide rates than does the US. There are widely available substitutes to guns everywhere.

Fourth, the authors make little comment themselves on the “rational” suicide theory: that for a person completing suicide (where little pretence and strategic posturing is involved) the benefit of living was judged to be worth less than the benefit of dying. Yet when we look at the single most powerful findings in the 3 sets of data, they are as follows:

  • the depression score was most highly significant for suicide attempts;
  • the divorce rate was most significant for county deviations in suicide rates;
  • when it came to looking at time variation itself, the rise in female participation had the most explanatory power for state data, and median income was the only other significant variable at the 96% level.

For future research, then, it will be interesting to explore the interaction there may between people with a predisposition to suicide, i.e. where clinical depression is present, and the pressures on families that arise from increasingly competitive economic environments. It may well be that increasing female participation in the workforce, divorce, delinquency, drug and alcohol abuse, etc, are themselves endogenous to the dynamics that drive the rising youth suicide rate. The median income variable found in the state data suicide time variation is also interesting in this context. For example, youths may feel rising stress from the more competitive school environment. Rising income disparities put more onus of doing well in school: i.e. to be winners in the process rather than the losers. Depressed students cope less well as the pressures increase over time.

The interaction between these sorts of pressures and the presence of clinical depression may be a cocktail that is increasingly lethal over time.