Extrapolation Games

The assertion is that there are five million "unsafe abortions" in Africa, six and a half million in South Central Asia and a total of nineteen million in the world's most undeveloped countries. Exactly how do these analysts come up with these astronomical estimates of abortions in these regions when there is so little real data? The following hypothetical example shows how abortion estimates can be dramatically altered when they are extrapolations of small sample sizes based on uncertain and possibly outmoded data.

Imagine it is 2002 and an international "reproductive health" agency would like to estimate the number of "unsafe" abortions in a developing country of 10 million inhabitants. The only data set available to re-searchers is a 1984 hospital study of abortion admissions from a village of 5,000 people.

Researchers have been assured by "local experts" (perhaps local "family planning" advocates working closely with the international agency) that this village is especially representative of the nation as a whole in terms of its population demographics, birth rates, abortion rates, and health care quality. Since the laws of the country in question protect unborn children from abortion, every abortion would automatically be considered "unsafe."

Now suppose a study reports that in 1984 10 women were admitted to the village hospital for complications from induced abortions. Where do we go from here?

10 Abortions in the Village, or 100 Abortions?

WHO starts with an assumption that from 10 to 50 percent of all abortions produce complications requiring medical care.

If 50 percent (half) of abortions require medical care because of complications - - and 10 women who aborted were admitted to hospitals - - our hypothetical data analyst would estimate that 20 unsafe abortions occurred in this village in 1984. But if only 10 percent (one in ten) of the total number of abortions require medical care, the formula says there were 100 abortions in that same village. Already, our estimate has exceeded our evidence by a factor of 10.

40,000 Abortions in the Country, or 200,000 Abortions?

Next, the agency calculates a ratio of abortions to live births. Suppose that births in this country in a given year are projected to equal 3 percent of the estimated current population, a percentage not unusual in the developing world. Three percent of our hypothetical village of 5,000 people would mean 150 babies. Three percent of 10 million would mean 300,000 births for the country as a whole.

If the agency goes with the lower abortion estimate - - that there were 20 abortions in the village - - our researchers will divide this abortion figure of 20 by the village's 150 live births. This results in a ratio of 13.3 abortions for every 100 live births. Applied to the country's expected number of births, that would mean a claim that 40,000 abortions took place in the nation.

But if the agency took the higher estimate - - 100 abortions - - it would come up with a far different ratio - - 67.7 abortions for every 100 live births (two abortions for every three live births). Applying that ratio to the entire country's estimated 300,000 births, the agency would then project 200,000 total abortions. All this from a report of 10 women having abortion complications in one small village.

Or Could That Be 800,000 Abortions?

Now imagine the agency believes (as WHO does) that many women entering the hospital who claim their complications are from miscarriages are actually suffering complications from abortions, but were afraid to say so. WHO's report does not specify the estimate for "underreporting" used in its calculations, but says that sometimes only 25 percent of abortions (one in four) are reported.

Suppose our data analyst assumes that women who stated induced abortion complications as their reason for hospital admission represented only a fourth of the women who actually were coming to the hospital because of abortion-related complications. If so, then he or she would estimate that not 10 women but 40 had actually sought treatment for botched abortions at the village hospital.

Repeating the same calculations done in the previous section but with a base number of 40 instead of 10, we develop an estimated ratio of as many 267 abortions for every 100 live births, or 800,000 abortions, performed in the country.

But Wait, We Could Have 933,333 Abortions!

Finally, suppose that the projected births for that country in a given year turned out to be a little bit off. Reflecting its pessimism about the country's ability to slow its population growth, the UN mistakenly estimates that births will equal 3.5 percent of the population, or 350,000.

Our hypothetical researchers now use this figure in their abortion estimations. If we assume a higher number of births (and stay with the ratio of 267 abortions for every 100 live births, as estimated above), we obtain an estimate of 933,333 abortions. This figure is more than 23 times greater than our original lowest estimate.

1 Million Abortions in the Region, or 23 Million Abortions?

Suppose our "reproductive health" agency now wants to estimate the total number of abortions performed in an entire region. According to raw data, adjusted only to cover the entire region's population, let us say that actual studies in countries throughout this region counted 1 million abortions.

However, if our researchers and their "local experts" believe that they need to make the same adjustments to other countries' data as they do to our sample country, the total number of abortions they end up reporting could then be 23 times higher than their original count. It could then be a matter of "experts' judgment" as to whether the region performed 1 million abortions, or 23 million abortions.

With the international "reproductive health" agency's funding riding on size of the crisis, which figure do you think its experts likely to lean toward?