Log in

No account? Create an account
An interesting editorial from the Wall Street Journal. However, I'm… - Silicon Rose [entries|archive|friends|userinfo]
Silicon Rose

[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

[May. 18th, 2006|09:32 am]
Silicon Rose
[Current Mood |thoughtfulthoughtful]

An interesting editorial from the Wall Street Journal.

However, I'm not so much going to comment on it, as on editorials and facts in general. I think one of the biggest problems facing us in our challenge to advance is the inability to find truths. In science, you can get a certain level of certainty about material things, though you are still subject to a certain amount of uncertainty as well. For example, you could come up with a false causality link that happens to hold up through a few confirming experiments, but fails later, throwing scientists who see the 'wrong' behavior into confusion.

But when you're dealing with soft sciences like parts of psychology, economics, etc. - well, frankly, anything having to do with people instead of physical laws - one of the biggest tools in your arsenal is statistics. Statistics seem to be a fundamentally flawed method of finding truth to me. They can say so many things. Perhaps when properly applied they are useful, but it's far too easy to misuse them, thus the phrase "lies, damned lies, and statistics."

Life is full of hard problems. Not all hard problems are easy to recognize, but there are a certain set of hard problems - education, fairness - that are. However, the accomplishment is not in finding hard problems, even ones that are difficult to find, it is in solving them.

Too close to a tautology to be useful? Probably. But it's far too easy to complain and not do anything. I'm as guilty of this as most people.

[User Picture]From: harinezumi
2006-05-18 05:25 pm (UTC)
Statistics are a genuinely useful tool when used properly. The problem is that the form in which they are usually presented to the general public makes them completely worthless. Things like sampling methodology, the way the information was collected from the sample, or even the source of a study's funding have tremendous importance in evaluating the validity of the resulting statistical data, let alone the analysis thereof.

There's also a fact that they drill into your head in your first Statistics course, and keep on drilling at every opportunity thereafter, and that is that correlation does not imply causation. In spite of this, a fair number of Stat students never quite manage to wrap their heads around this fact, if my lab partners were any indication. It's no wonder then that laymen are easily swayed by misrepresented data and demagogues are all to happy to misrepresent it.
(Reply) (Thread)
[User Picture]From: froborr
2006-05-18 07:47 pm (UTC)
I would argue that, given its disdain for making testable predictions, economics is more of a mathematized philosophy than a soft science, though that distinction is a spectrum, not a hard and fast line.

But you are correct about how easy it is to misuse statistics. The editorial you linked, for example, talks about income growth averaged across all citizens, without breaking it down by economic class. It thus utterly fails to answer the primary claim of opponents of supply-side tax policy, which is that it encourages a form of economic growth which benefits a wealthy minority while ignoring or even actively harming the majority of citizens. The statistic about job growth is equally specious in isolation: it is useless to double the number of jobs if you simultaneously raise the cost of living so much that everybody needs to work two jobs!

Finally, I would argue that science does not provide certainty at all. Certainty is an emotional state, unrelated to the truth value of the proposition in question, and science is a method of institutionalized, perpetual doubt. Indeed, I would argue that science's greatest strength is that it considers certainty undesirable, and instead pursues correlation to observed reality.
(Reply) (Thread)
[User Picture]From: cyfis
2006-05-20 06:14 am (UTC)
Previous posters have more or less summarized the primary problems with statistics as reliable method of gathering and analysis (particularly the latter) of information, so two cents on the article content instead. Encouraging economic growth through reducing taxes may be effective, but without some closer research into the various industries or, as froborr was pointing out, the economic classes who stand to benefit from these reductions, it's a lot more difficult to predict the impact of the tax reductions on any particular demographic, or whether or it benefits the country as a whole, by whatever yardstick you choose to measure that.

Unfortunately, while it is possible to misuse statistical data, there really aren't that many other methods for gathering information in the fields of social science. Asking whether someone is satisfied with their employment and social standing does not have a definite answer such as the speed of light or whether a particular element is reactive to oxygen.
(Reply) (Thread)