The Romanian Baby Bump
UPDATE 11 SEPT 2012: Gapminder has recently updated their birth statistics for Romania to be more accurate.
Steven Levitt and Stephen J. Dubner, the authors of Freakonomics, used the example of Romania’s 1966 sudden ban on abortion in support of their controversial theory of the connection between abortion and crime rates.
Obviously, since immediately before the ban took effect, some Romanian women were pregnant who would have otherwise terminated their pregnancies, the birth rate (or, a similar concept called the ‘Total Fertility Rate’) should have jumped up in 1967 (and beyond, depending on how women responded to the change in policy.)
I wanted to use Dr. Hans Rosling’s Gapminder software to see the effect of this ban on birth rates, crime rates and other social metrics, but noticed something strange in the software’s data files.
The graph below shows Gapminder’s values for the Total Fertility Rate in Romania from 1960-2006 as a blue line. It’s interesting to see Romania’s birth rate decline precipitously in the early 1990’s as communism (and large sections of Romania’s economy) collapsed. Events like this give some credence to the idea that children are “normal goods” (that is, that people tend to want more children, all else held constant, as they get richer – and vice-versa). The US and Western Europe saw similar drops in fertility rates during the Great Depression.
(click image to enlarge)
But what’s just as interesting is what’s not present: a spike in the TFR in 1967. Instead, the Gapminder’s data just has the fertility rate rising smoothly throughout most of the 1960’s.
The black line is taken from the Stockholm University report ‘The Standardized TFR‘ from Jan M. Hoem and Cornelia Muresan.
Overall, the basic trends in Gapminder’s data seems to be very similar to the data of Hoem and Muresan, but perhaps Gapminder is using some sort of smoothing to eliminate things like the 1967 fertility spike or the smaller drop around 1984. (Gapminder I noticed, also tends to do things like assume a given value is constant or to linearly interpolate between two existing values when data is not available for certain years.)
The percent differences are quite sizable – Just after 1965, for example, Hoem and Muresan’s data indicate that the TFR was slightly less than 2.0 while Gapminder says it was over 2.5, a difference of more than 25%.
More importantly to me, so-called ‘step change’ are very interesting (like the sudden abortion ban, or various stock and commodity market ‘flash crashes’ that have happened in recent years). And, unfortunately, Gapminder seems to have smoothed out this baby bump.
I think it would be helpful if Gapminder Desktop, the downloadable version of their software, had a feature that allowed users to upload their own data, so that I could at least see the alternative TFR estimates in my own copy of Gapminder.