The Scottish philosopher David Hume (1711-1776) understood a bit about both causation and epistemology (for the latter, meaning the manner with which human beings gain knowledge and understanding). He made a very simple observation, but one that was radical for its times: Just because A and B appear together does not mean any inference can be drawn that either A causes B or that B causes A, though the experience of A appearing with B often yields the expectation in the minds of observers that B will appear when A appears. All we can really know, he said, is that each time we have thus far seen A, we have also happened to see B. He understood the human proclivity to seek causes for effects, and that the resultant perceptual bias often deceives us into thinking we understand a thing’s causes and we actually don’t. But he went further than just pointing out that correlation is not causation to claim that causes are unknowable; that what we see as cause and effect relationships are just a mental conflation of two phenomena or objects appearing together.
According to British philosopher Bertrand Russell, writing in the early 1940’s, Hume’s skepticism has never quite been satisfactorily answered, from Russell’s, “The History of Western Philosophy”:
To refute him has been, ever since he wrote, a favorite pastime among metaphysicians. For my part, I find none of their refutations convincing; nevertheless, I cannot but hope that something less skeptical than Hume’s system may be discoverable.
Since nothing less skeptical could answer Hume’s observation—there really is no way to know with absolute certainty much of anything—Hume was mainly ignored. The Industrial Revolution, whose technological developments depended heavily upon understanding cause and effect relationships, no matter how ultimately doubtful, moved swiftly along on the rails of empiricism, which taught that multitudinous observations could be inductively employed to flesh out reliable and useful cause and effect relationships. If A’s appearance was always followed by the appearance of B, and B never appeared without A, it was reliably clear that either A was causing B, or some third factor was causing them both.
Without answering Hume’s skepticism, the intellectual pendulum swung away from Hume to empiricism. It may well be that nothing of causation, or of really anything, is certain, as Hume claimed, but once enough observations are accumulated, empiricism provides that some reliable and useful conclusions about causation can be formed. Empiricism—understanding based on observations—has worked so well until the intellectual pendulum now seems today to be swinging almost to mysticism, where not even observations are necessary to claim the existence of causal relationships, or where those observations that occur simultaneously are immediately imbued with causative inference, no matter how fortuitous their simultaneous appearance might otherwise be. As the following three examples illustrate, some measure of Hume’s skepticism might be properly redeployed today to restrain the galloping hubris of empiricism. I stumbled across each them in my reading this past weekend.
The first example comes courtesy of an article in The Economist newspaper. Its headline tells it all, Baby Monitor: In poor countries, lower fertility is usually good for growth. But it can also lead to inequality. There are at least two unproved suppositions here—that in poor countries, lower fertility is good for growth, and further, that lower fertility can lead to inequality. Even for an empiricist, there is no reason to conclude lower fertility is causing either growth in poor countries or inequality. Just because two phenomena appear together does not an inference of causation make, and more poignantly, even if two phenomena have some sort of causal relationship, their simultaneous appearance is not dispositive of the direction in which causation runs.
It may very well be that growth in poor countries causes declining fertility. It also may be that growth causes inequality. Which seems more reasonable, to imagine that women quit having babies in order that aggregate economic growth and development—a phenomenon completely abstract and foreign to their decision to have babies–might obtain? Or, perhaps, that aggregate economic growth and development has changed the cost-benefit calculus for how many babies the average woman wishes to bear?
The decision or not to have a baby is exceedingly personal to the person whose body will suffer the pains of gestation and childbirth. Even before the pill and abortion, even in supposed backward countries that recognize little in the way of female rights, women have always been the ultimate arbiters of the rate at which children would be born. Of course, Hume would say, even if true, that doesn’t mean women will continue to make those decisions, but empirically speaking, there is really no way to ignore the woman’s role in deciding or not upon childbirth. And to imagine (like The Economist and many other social commentators) that the availability of modern contraceptives (like condoms) might determine the rate at which children are born in poor countries is to imagine that women in poor countries are complete imbeciles, not much different in their procreative planning than are rabbits. The evidence suggests otherwise.
It is well-documented, that in every last developed country, female fertility has plummeted with economic development. (In China, it plummeted even before much of the country enjoyed development, presumably the result of its draconian one-child policy, which raised the cost of a second child to an extreme level). Excepting China as an outlier, which came first, development, or fertility declines? If it is assumed that women are the final arbiters of the number of babies their bodies will bear, it must be that somewhere along the development continuum, women decide there are other and better things to do with their lives than simply becoming baby factories.
What happens economically that might change the calculus? First, as economic development proceeds, agriculture becomes industrialized, and the portion of the population employed in agriculture plummets. People move from the farm to the cities, and no longer need seven or eight children per family to ensure a few survive to help in working the land, which brings up another critical aspect of development. While the need for babies declines with urbanization, improved access to medical care and immunizations means a great many more children survive into adulthood with economic development. Fertility rates (the number of children born to each woman) and infant mortality rates are closely correlated. As infant mortality rates decline along with economic development, fertility rates decline along with them. And this has happened without condoms, pills or abortion, as the phenomenon arose in the West and Japan long before such birth control conveniences arose.
But the truth is, we don’t know exactly why economic development and declining fertility rates appear together, we just know that they do. When the “A” of economic development appears, the “B” of declining fertility rates does also. A colorable argument can be made that since economic systems don’t decide or not to make babies (except as in China, in putting an upper limit on their production per family), it seems more reasonable to view the phenomenon from the perspective of the individuals who do. But to imagine, as did The Economist headline, that lower fertility causes growth (or income inequality) is to draw an inference of causation without justification. Even if we abandon Hume, and believe with the empiricists that causation is ascertainable, we still don’t get to render conclusions of causation on simply the correlation of two emergent quantities which themselves exist mainly as the abstract detritus of a chain of causal sequences.
Empiricism, never mind Hume, is often rejected by medical practitioners and patients. As an article in the New York Times makes clear, a goodly portion of what are standard procedures in treating disease have no rational basis justifying their performance, from Testing What We Think We Know by Dr. H. Gilbert Welch:
BY 1990, many doctors were recommending hormone replacement therapy to healthy middle-aged women and P.S.A. screening for prostate cancer to older men. Both interventions had become standard medical practice.
But in 2002, a randomized trial showed that preventive hormone replacement caused more problems (more heart disease and breast cancer) than it solved (fewer hip fractures and colon cancer). Then, in 2009, trials showed that P.S.A. screening led to many unnecessary surgeries and had a dubious effect on prostate cancer deaths.
In medicine, where a person’s life might be at stake, the impulse is to do anything that it is imagined might improve a patient’s odds, even things that have generally proved deleterious to patient health, such as Welch’s examples of hormone replacement therapy and PSA screening. Welch also lists mammograms and colonoscopy as suspect procedures that possibly impose costs well out of proportion to anticipated benefits.
Even without considering the monetary cost of these procedures, it may well be that outcome-based medicine, i.e., empirical medicine, would counsel their abandonment. The first rule of medicine (which should really also be the first rule of life) is to do no harm. These standard procedures, if the studies can be replicated, violate that first precept of medicine, and ignore the first foundation of empirical observation—that correlation, while not determinative of causation, is a necessary predicate. With no improvement in outcomes, there can’t even be a correlative reason from which causation might flow to justify conducting the procedures. Forget Hume, even the empiricists would say the thinking here is utterly baneful. Spending money and pain on procedures that aren’t expected to bring anything but money to the doctors and pain to the patient? We’d do better to tarry down the rabbit hole of Hume’s skepticism than go along with nonsense like this.
The final example comes from the “science” of climate change. In the cacophony of global warming alarms over a really warm July this year, you may have missed a small data point, one that folks like James Hansen, the gadfly global warming polemicist with a government job at NASA, would surely like to prevent from gaining widespread dissemination. Carbon emissions in the US this year declined to levels not seen in twenty years, mainly because natural gas got so cheap that it was switched for coal in a number of electricity-generating plants across the US. (See article).
Thus Hume was right after all. Even if every time A appears, B also appears, we can’t thereby deduce that B will always appear when A appears, because we can’t really know what, aside from correlative appearance, is the relationship between A and B. If carbon dioxide levels in the US decline (A), but B still appears (the hottest July on record), then it would seem wise to have foregone the empirical induction that the appearance of A causes B. Of course, I am engaging in a flight of logical fancy–nothing of this holds any empirical validity whatsoever. Carbon dioxide in the ocean of the atmosphere is hardly barred from entry to US skies (though the only audience with which Mr. Hansen seems concerned is the US), and hot temperatures in the US hold precious few implications as to the climate of the whole planet. But it points to the outrageous fallacies in causative analysis underlying the anthropogenic global warming catechism. With global warming, the intellectual pendulum swung from Hume right through empiricism clear over to mysticism, which is to say, right back to Hume, where causes and effects are unascertainable.
Causation analysis is hard, so hard, Hume would say, that it is impossible. And although Hume’s assertion that causation is impossible to definitively prove has never been satisfactorily refuted, it would be impossible to survive in the world of Hume’s imagination (a challenge which Hume himself refused to engage). Reliable and useful causative inferences have repeatedly and successfully been drawn from empirical observations. The idea that effects have causes forms the foundational premise of pretty much every development of mankind since the beginning of time. But it is right and proper to be mindful of Hume’s skepticism when evaluating claims of causation. Just because A and B appear together, without more, we then only know that A and B happen to have appeared together. A and B might have a causal connection with each, or may have just fortuitously appeared together, or might have a causal connection to some third event or condition. In complex systems, causation might be a confluence of innumerable factors, making them unascertainable as a practical matter. What causes a hurricane? What forces societal disintegration? Which dagger killed Caesar? Even if a causal connection is established (to an empiricist’s satisfaction), it remains to ascertain in which direction it runs. Does A cause B or B cause A? Causation is hard, not easy. It takes humility, not hubris to succeed in teasing out causative relationships.