Correlations May Not Prove the Cause

Chapter index in this window —   — Chapter index in separate window

This material (including images) is copyrighted!. See my copyright notice for fair use practices.

Often the observation of a correlation between two observables is used to proclaim a cause-effect relationship between them. For example, suppose that there was a possible correlation between sex education in schools and a recent rise in venereal disease and teenage pregnancy. One could say that sex education has caused the rise in VD and teen pregnancy, but the scientist cannot say that without a more detailed investigation.

After all, there are many other factors that could be the causal agent behind this problem. A rise in the population of teenagers is possible, causing every activity related to teenagers to go up: automobile accidents or purchasing particular types of clothing and albums. Few would claim that sex education in schools has been the cause of increased purchases of acne lotion. There could be an increase in the population of particular types of teenagers, those in an area of the country where sex education is not taught or where early sexual experimentation is encouraged by various social or family pressures. There are many variables possible to produce that correlation. Correlation does not prove causation. A correlation between sex education and teen sex problems does not prove a causal connection, and, by itself, it does not give us a clear indication in which direction there may be a connection. For all we know at this point, an increase in teen sex problems has led to an increase in sex education classes!

Another example is the correlation between smoking and lung cancer occurrences. After a couple of decades of study the government decided in the 1970s that there was a causal connection between smoking and lung cancer and changed the warning label from ``Caution, smoking may be hazardous to your health'' to ``Caution, smoking is hazardous to your health''. A 1950s study only controlled the basic environmental variable-lung cancer for smokers living in the cities vs. lung cancer for smokers living in the country. This study was roundly criticized and rightly so. There were many other important factors that needed to be looked at such as diet, healthy or unhealthy occupations, stressful occupations, or genetic factors.

By the 1970s, more careful studies each incorporating tighter and tighter controls based on possible oversights of the previous studies had proven to the government's satisfaction the causal connection between smoking and lung cancer. By the 1980s other diverse corroborating factors had been identified-from the effects of secondhand smoke to chemical analysis of cigarette smoke revealing over 200 toxic substances, including radioactivity.

Despite all of this study, we really cannot say that cigarette smoking has been proven to be the principal cause of lung cancer. A scientific proof is not known with absolute logical certainty. A controlled study can never be completely controlled-there are just too many possible variables. The link between smoking and lung cancer cannot be known in the sense of ``known beyond any logical or conceivable doubt.'' The point is, however, can we say we know that cigarette smoking is a principal cause of lung cancer beyond a ``reasonable doubt''? Is it rational if we claim to know something even if we are not absolutely sure that we know something? Can we distinguish between what is ``conceivably'' true and what is ``reasonably'' true?

A humorous example of the difference between a correlation and a cause-effect relationship is the Coalition to ban Dihydrogen Monoxide. To find out more about this ``dangerous'' chemical, select the links below (will display in another window):

  1. Ban Dihydrogen Monoxide!
  2. Dihydrogen Monoxide Research Division.


previousGo back to previous section -- next Go to next section

Go to Astronomy Notes home

last updated: June 1, 2007

Is this page a copy of Strobel's Astronomy Notes?

Author of original content: Nick Strobel