New York City’s Health Department (motto: It’s for your own good!) released its annual birth and death statistics, and the internet is aflutter, particularly over the numbers on abortion. “Raw” statistics show that 2 out of every 5 pregnancies end with procedures which kill these females’ unwanted fetuses.
I use this language, incidentally, because it is the least euphemistic. Perhaps it is not the pleasantest way—softening truths is the very point of euphemism—but it is the plainest. Let’s not discuss here whether it is moral for a female to kill her fetus (or whether she should have that job hired out). Humans have been killing off fetuses, infants, toddlers, teens, adults, and the aged for all of history. Let’s just not ask whether any of these recipients of homicide had it coming: we will not come to a satisfactory answer. Let’s instead look just at the numbers.
Most people will skip over the Health Department’s technical appendix. A mistake. That appendix explains why the numbers in the report are not really the right numbers, of how difficult it is to count accurately, about the flawed procedures in collection, about how, even, that some numbers are the output of statistical models. This means that whatever use we make of the (approximate) numbers will come attached with a goodly amount of uncertainty; further, an uncertainty that is difficult or impossible to quantify. This goes double if we input the Health Department’s numbers into statistical models of our own, which only can increase uncertainty. Be wary of anybody’s firm conclusions.
The most obvious kind of error is non-reporting. “Spontaneous Terminations of Pregnancy” is tallied; but this number is only from those pregnancies reported to the Health Department. Of those women whose pregnancies end in, say, miscarriage before these women visit a physician, we know nothing. The “Borough of Residence” is given for pregnancies, but don’t know how many people lied, say. We keep these uncertainties in mind when we read that, in 2009, there were 225,667 (tracked) pregnancies, of which about 5% where spontaneously terminated, and of which about 40% (2 out of 5) terminated non-spontaneously (I’ll let this euphemism pass).
Nearly 4 out of 5 pregnancies are terminated non-spontaneously when the mothers (or mothers-not-to-be) were under 15 years old. This falls steadily to about 1 in 4 when the mothers-not-to-be are 35-39. Notice that age buckets in the picture are not constant; and that the “jump” for ages greater or equal to 40 is an artifact of very large bucket. Measurement error will certainly shift these points higher or lower, but the direction of the trend is probably accurate. We cannot, of course, say why from these data alone.
Abortion has been used for male sex selection (e.g. China). Is there evidence of this in New York City? Not so much. The under 15 age bucket only saw 112 live births, so not too much can be made of the low percentage of male babies in this group. And the 15-17 group only saw 2,308 births. There is still a bucket effect for the 40 and older group; but it’s difficult to say which direction the overestimate occurs. Overall, 51.3% babies were boys, a typical figure. We cannot say anything about any particular abortion, but on average, there is not good evidence that male sex selection is occurring.
Marital status (difficult to measure, or so the appendix says) is correlated with abortions. The percent of births to married mothers increases steadily as moms age.
New York City makes it next to impossible to discover the correlation of race on abortion. Forms have to be filled out for each abortion, but they do “not contain the woman’s name or identifying information.” They must contain more information than this, because there are breakdowns by age, and one partial statistic by race, the city just don’t report them. Abortions for ages 15-19, and only these ages, are given by race. Asians 72%; Blacks, 72%; Whites, 64%; Hispanics, 52%.
Except to note that the total number of abortions in the city have decreased slightly through time (as have all births), and that it’s possible to break some numbers down by borough, there’s not much more of interest we can say (about just the numbers).
It’s not easy being teen nowadays. The graphs tell me that our needs, priorities, views and perspectives about life and how we deal with the societal pressure do change as we grow older, and that those changes seem to be for the better.
Statistical models are used to study and to quantify the uncertainty in data. Whether they are useful or good is another story, however, they don’t increase the uncertainty in data. For example, a statistical model has been employed when calculating the sample percentages in this post, and it hasn’t increase the uncertainty. Perhaps, this statement is meant to say “garbage in garbage out.â€
Not all euphemism is propaganda. It is also a marker of a polite person. Statistical thinking is important and not an easy thing for people to do. IMHO, you make it harder for some to benefit from your blog by being impolite.
It is surprising to me that more people are not aware that abortion is tantamount to genocide of the unwanted races. I suggest that readers research “black genocide” and the racist eugenic roots of Planned Parenthood.
If “diversity” is the PC goal, then why is it PC to kill 70%+ of all pre-born black babies?
PS — please forgive my impolitic rhetoric. It is not my intention to offend anyone. Just asking pertinent questions.
George,
What exactly is your objection?
JH,
Models do not increase uncertainty in the data, but they add uncertainty to conclusions. Models are always assumed true.
@Mike D.
Babies are aborted because their mothers don’t want them at that time. Mothers are generally from the same race as their babies.
Genocide means killing of as many people of a race as possible, irrespective of age. AFAIK, most mothers leave the abortion clinic alive.
Sander,
Thank you for that clarification. That explains and forgives everything. Hooray for abortion — it’s not genocide. More power to the dead baby “clinics”.
BTW, darn those stupid, backwards, religious nut cases who think fetuses are worthwhile human lives. Just goes to show how philosophically bankrupt Christianity is.
The abortion rate in NYC is down from a decade ago. There’s some hope.
Mr. Briggs,
Models add uncertainty to conclusions?! I guess some economists would agree with you that there is the extra uncertainty of not knowing whether the model is true.
When assessing the uncertainties in the parameters of a statistical model, there is no other choice but to assume the underlying model is true. This is not to say that it’s the true model.
Measurement errors can add uncertainty to conclusions, so can contradictory findings. I know I can add uncertainty to a model by introducing an error term based on the data structure.
However, that models add uncertainty to conclusion is a new concept to me! I’ll need more explanations or pounding to get it in my head. In fact, it’s a bit counter intuitive to me.
Statistical concepts can be closely reflected in life. (Which is one of the things I like about statistics.) To reach a conclusion with less risk/uncertainty, one would sort through information, i.e., model the data, to find patterns and evaluate variability.
And statistical modeling or an appropriate model provides us a way to decipher the uncertainty in data and hopefully minimize the uncertainty in conclusions. Not an easy task though.
Data which is spit out of a model is more uncertain than raw data. And, if the data taken from the model needs further “refinement” (i.e. it doesn’t tell us anything useful), putting it another model adds more uncertainty. Conclusions made by models of models are, of course, more uncertain than raw data which has obvious trends.
Correct me if I misunderstood anything.
Adam,
A model is used to make predictions and doesn’t produce any data. The prediction uncertainty from a model heavily depends on the ascertained data (and also the goodness of the model, of course). For example, I’d like to predict the exam1 score of Matt who misses the exam and will take it on a different day. It’s observed that the rest of 94 students in the class ALL scored 80%. I’d predict that Matt would score 80%. I won’t say there’ll be no prediction error. However, if he doesn’t score 80%, I would probably conclude that something out of ordinary has happened.
If one uses, say, temperatures re-constructed from a model, the problem is not the model add uncertainty in conclusions or predictions. It’s the fact that the reconstructed temperatures are not observed/true measurements of temperatures. There is an error or deviation between the reconstructed and the true values. And it’s the error that causes the extra uncertainty, not the model. Poor quality data (though the re-constructed temperatures really are not “dataâ€) make bad predictions.
I hope my explanation helps.