Economics is often called a social science. But is it really a science? Science considers data from the real world in order to ascertain truth. But analysis of data alone is not enough. The way the data are analyzed is very important. It needs to be considered in an objective way with a critical eye. This is not always easy to do.
Disciplines that are undoubtedly scientific in their approach can still be practiced by individuals in a nonscientific way. It is very difficult, in many cases, to approach issues in an objective and unbiased fashion. However well intentioned, those who consider only the data that support their biases cannot be called scientists. A better term might be "advocates."
At its best, economics strives to be scientific and inquire into the nature of the economy and society using the scientific method. In the social sciences, this is not always easy.
The scientific method involves a continual iteration between objectively observing the real world and formulating testable hypotheses that might explain these observations. In hard sciences like physics or chemistry, objective observation involves controlled experiments. Experiments are conducted in a controlled environment and are ideally replicable by other experimenters. If another scientist replicates the environment and conducts the same experiment, he or she should get the same results.
Clearly, this is not always possible in the social sciences. For example, if I were to approach the U.S. Federal Reserve and propose an experiment in which it doubled the U.S. money supply overnight so I could observe the effects on the prices of goods and services, they would rightly treat me as a crackpot.
Even if I were somehow able to convince the Fed to do so, the experiment would not be controlled because it would not be possible to replicate it at a later date with the same environment, since economic conditions unrelated to the money supply would undoubtedly change in the meantime.
To analyze the economy, economists must rely on "natural" experiments in which some key variable changes. This requires heavy use of statistical techniques.
For example, I might choose to look at episodes of rapid money growth over time and across various countries. In so doing, however, other variables that also influence the price of goods and services would not remain constant.
As an objective researcher I would need to ascertain whether the price movement I do observe can reasonably be said to be caused by rapid increases in the supply of money, rather than by some other factor. That is, I would be testing a hypothesis using statistical tools and techniques rather than by holding all other potential causes constant, as I would in a controlled experiment, which would be preferable but impossible.
For this reason, the economics profession puts a great deal of time and effort into training economists in the best available statistical techniques. Indeed, a large and important field in economics is econometrics, and many professional economists dedicate their entire careers to exploring the best ways to test various hypotheses for various types of economic data.
One drawback from natural experiments is that there is always some underlying randomness that is difficult to purge. When a test is performed it is done with some level of confidence, typically 95 or 99 percent. We reject a hypothesis if the chances of getting a result randomly are 95 percent, for example. Note, however, that this means on average we will reject a true hypothesis 5 percent of the time.
In practical terms this means that the determination of the truth of a hypothesis requires testing with more than one data set and often with more than one method of analysis.
At least in terms of hypothesis testing, economics strives to be scientific. What about how the hypotheses are formed in the first place? That is the realm of economic theory, and I will save that discussion for my next article.