What is the suicide
rate in Bolivia? Indonesia? Kenya? No one knows because government
officials in those countries don’t collect or report official suicide statistics. In fact, more than 100 nations don’t report suicide statistics to international organizations such as the World Health
The lack of reporting is unfortunate because suicidal behavior is a public health problem in some countries, and statistics are useful to government planners, public health officials, and medical researchers. When official statistics are unavailable, estimates can be used in their place. But how can officials estimate suicide rates?
Several years ago, Zorel Zambrano and I developed a method to estimate suicide rates in nations that don’t report official suicide statistics. The strategy is conceptually simple. First, identify many indicators (like the divorce rate) that are correlated with national suicide rates. Second, use a mathematical procedure called multiple regression to identify the best predictors of suicide rates and then generate a prediction formula. Third, use the formula to estimate the suicide rate in a nation, assuming one can find national data for the predictors in the equation.
In our study we included 73 nations for which suicide data are available from the World Health Organization. Most of the nations are in the Americas, Europe, Central Asia, and the Pacific. Only a few are located in Africa, Southeast Asia, and the Middle East.
For each nation, we recorded the average suicide rate for the years 1992-2003. We also recorded, when possible, values for many indicators that researchers have identified as moderate or strong predictors of national suicide rates. These indicators include divorce rate, fertility rate, adult literacy rate, GDP per capita, individualism rank, and locus of control. (Locus of control or LOC is a generalized belief in one’s ability to control the things that happen in one’s life).
Using standard regression techniques, we isolated the four best indicators of national suicide rates—divorce rate, locus of control, per capita GDP, and fertility rate—and generated three different prediction formulas, each with its own advantages and disadvantages (see Zambrano & White, 2009). One of the formulas, which can be used only when the divorce rate and average LOC are known, accounted for 79 percent of the variation in suicide rates, making it a powerful tool for estimating a nation’s suicide rate.
To test the accuracy of the formula, we conducted an analysis of suicides in Australia in 2003 (because the WHO had not reported Australia’s 2003 suicide statistics at the time of our study). Social scientists might expect Australia’s suicide rate to be higher than average, given its relative wealth and low religiosity. Our formula, however, estimated the suicide rate in Australia in 2003 to be 12.30 per 100,000 persons, which would be lower than the global average.[i]
We calculated the number of suicides by searching Australian newspapers on-line for government news about suicide statistics. According to several sources, 2213 persons committed suicide in Australia in 2003. Given Australia’s 2003 population of 19.86 million, we calculated a rate of 11.14 suicides per 100,000 persons—a number remarkably close to our estimate of 12.30.
Is our estimation formula useful? We believe so. Suicidal behavior is a major problem in some parts of the world. Public health officials are eager to know the number of suicides that occur each year. Unfortunately, some countries don’t collect official statistics. When knowing is not possible, estimating is the next best thing.
Portions of this post were published previously in:
Zambrano, Z., & White, L. T. (2009). Estimating suicide rates in nations that do not report suicide statistics. Undergraduate Research Journal for the Human Sciences, 8. Published on-line at http://www.kon.org/urc/v8/zambrano.html.
[i] In our sample of 73 nations, suicide rates ranged from a low of 0.75 (Azerbaijan) to a high of 44.85 (Lithuania) per 100,000 persons per year, with a mean of 13.26 and standard deviation of 9.96.