We've heard a lot of economics-bashing and a lot of bashing of the social sciences in general. They are supposedly useless.
Are they? Studying one, I admit they certainly have their pitfalls. But are they useless? Far from it, as long as we know their limitations. So I thought it would be useful to examine the social sciences, their succeses, and failures, and the reasons for them.
1. Universalia ante res an post rebus?
Let me tell you a story. Once upon a time a man called Immanuel Kant asked himself how man can reach eternally valid conclusions. To make a long story short, he concluded that, because of the limits of human ontology, any universal conclusion must be formal in nature: Mathematical. Any science is eternal as long as it's mathematically based. Once you start adding the material in, you use a posteriori data and necessarily relativise your conclusion.
This was groundbreaking for law, since Kant destroyed the position of rationalistic natural lawyers (and all proponents of natural law): They used a set of axioms, from which they tried to deduce an eternal legal order; those axioms were a posteriori, deduced from specific societies and of course not eternal: The limits of culture were the limits of universality. In fact, the error was the same the scholastics made: They, too, used axioms from which they tried to deduce an eternal legal order. The source of the axioms was different, religious, the presumption identical (In fact, most rationalists up to Leibniz were scholastically trained, which explains their method.).
In short, in society the best we can do with regard to eternal truths is a formal relation - and even those conclusions are often dependent on culture and social relations. For example, criminology tells us that poverty is one of the genrators of criminality. That claim may well be universal. However, all this means is that a = - bx +/- z, and all data which fleshes this out is transitory (a is the level of crime, b is the level of poverty, z is a dummy variable representing free will, missed data, and chaos in general). How much and how poverty influences criminality varies.
In short, while I may be able to deduce correlation and even causation, I won't be able to foretell exactly how crime will increase when poverty increases.
Granted, the situation isn't as relative as this makes it seem. Humans, free will aside, are bound by the structures in which they find themselves. For example, let's say I live in Oregon, want to get to Washington in, at most, a week, and have $1000 in my pocket. My choices are: Car, bus, train, plane. Despite all that free will, my behavour is predictable; if I am in a hurry, I will choose the plane. Therefore, free will aside, my behaviour will be predictable, while allowing room for irrationality, and creativity, of course. That makes temporary conclusions more accurate.
And this bring us to economics, the "dismal science." Sociology deals with society and its social groups as a whole, as one might observe a coral reef, law structures society and both reduces the chance of and resolves conflicts, and economics, primarily concerned with the allocation of scarce resources, focusses mostly on individual activity (which is why macro is a late newcomer to the science).
The problem with any social science, as I've pointed out above, with the example of the person who wants to get from Oregon to Washington, is that any model, any prediction, simple or sophisticated, depends on certain assumptions, which allow us to collapse complex phenomena into a model or a conclusion. The catch is that no assumption is true.
For example, the assumptions of current state of the art economic models are [Frey, 1992]:
(1) Methodological individualism
(2) Consistent responses to incentives
(3) Self-seeking (egotism)
(4) Definable sets of preferences and constraints
(5) Bounded rationality, largely constraint by institutions
Each and every one of these assumptions is false. Humans don't consistently react to incentives. Not all people are egotists and people aren't always egotists. People may not have completely definable set of preferences, and they might not always be rational. Clearly, the models are worthless! Right?
Wrong. While I didn't say the assumptions are completely true I didn't say they are completly untrue, either. People don't respond consistently to incentives - but how consistent are they, in general? People are sometimes altruistic - how often and at what cost? How irrational can people become under some conditions?
The point is that no model is perfectly accurate. It may be true enough. Despite its failings, it might suffice for prediction. Simply put, if enough people chiefly act like egotists, the few who are altruists won't substantially alter predictions. This also shows how models can be improved, by refining assumptions. Note that the complexity of the model also depends on the task you need it for.
For example, older economic models assume substantive rationality. That means that an individual maximises his subjective utility tout court. And, if you're analysing the market dynamics of a town fair, that is accurate enough.
However, with more complex relations that assumption is shown to be too inaccurate to be upheld. One reason are transaction costs: The cost of the effort required to find a better choice is greater than the gain from it (I won't drive 50 miles to WalMart to save 50 cents on a salad). So they have to be included (in a fair, where everything is close, transaction costs are largely irrelevant). Then we see that people don't search for objectively optimal solutions, but ones which are good enough: An individual ends his search when he finds a solution that satisfies his aspiration level (bounded rationality). These assumptions are already far closer to reality than our original model. Not perfect, of course - no model can be perfect - but with better predictive capacity.
Sometimes a model purposely constructs an ideal situation which cannot be found in reality. An example is Telser's  work on self-executing contracts. Telser is a neoliberal economist and a formalist, which means he focusses on models. In his work he defines circumstances in which a long-term contract would enforce itself without outside intervention, without courts - law without laws, law without judges. We know that his assumptions aren't true. We know that such a situation is not found in reality. But his study isn't worthless.
The reason is that by reference to that ethereal optimum we can create incentive structures which lessen the chance of conflict and litigation and ensure the actors act as they are supposed to within the limits of the legal order. Telser's work is also interesting because it is irrelevant for practical purposes if one of his assumptions is false: That men are good and adhere to contracts only when the perceived advantage of doing so is greater than the advantage of not doing so - in short, as long as it is useful to follow the contract. Now, this assumption is useful in creating incentive structure because, if true of the other party, you will need it. If it isn't, then the incentive structure will be redundant since the other party will do the right thing for virtue instead of incentive. You've lost nothing (and people are a sorry lot so it's better to plan for a bastard).
In conclusion, all models are necessarily flawed, be they legal, sociological, or economic - or models in natural sciences. They are not true in the sense of being perfectly precise - but they may be true enough. And when a model fails it simply means you have to look or reasons why. But we know that doesn't happen. Why?
2. The Difference between Positive and Normative
A persistent problem of the social sciences is the difference between the objective and subjective. The problem can be illustrated with an example: We know that the gap between the rich and the poor is widening. Is that a social problem?
Objectivists would argue tout court that it is or that it isn't. This is the ordinary lay reference towards reality: Certain phenomena are (un)desirable in and of themselves and should be resolved.
Such a view empirically doesn't hold. Certain objectively harmful phenomena never attract the public's attention (smog), while other, objectively less harmful or even harmless phenomena become the crux of the public's obsession (texting while driving, paedophilia, the purported autism-vaccine link...). More frustratingly still, even when people agree that something is a social problem, solutions differ, and may be completely contrary.
As an example, the AIDS pandemic is a social problem. At least, most people agree it is so. Most people agree that AIDS is a problem of prevention and securing treatment for the infected.
However, not all do. A Camorra clan in Italy is famous for the fact that it keeps its brothels AIDS free to such an extent that clients do not need condoms. It does not do it by charity. Unlike most Camorra clans, which ensure their members have access to the bes doctors and state-of-the-art treatment if they get infected (it can be a 1caring profession at times) that clan does not. If they found out that someone, anyone in their territory is HIV-positive, they kill him, based on the logic that the only way to stop an epidemic is to cull the herd (If you are HIV-positive, consider yourself warned if travelling in that Camorra clan's territory).
This cruel example illustrates the difference between the objective and the subjective: We can agree on objective reality with a greater or lesser degree of certainty. But solutions are different. Solutions have different effects, and we can study them with a greater or lesser degree of certainty. However, the choice of solutions is a matter of goals, of taste, as is the question what a problem is. All phenomena need to go through a definitial process to become social problems.
In short, what is does not necessarily imply what ought.
This is something that economics often forgets: It creates optimums, mathematical or otherwise, and states which are supposely desirable, when what ought to be is inexorably relative.
An example is regulation. Regulation often increases prices. For example, European-style regulation of residential leases can in fact lower investment in real estate and contribute to a housing shortage. By increasing costs (an example are consumer loans) it can also price some out of the market.
So, deregulate! Err... why? We can discuss the objective parts of these problems: How much is investment reduced? How many people are priced out? However, deregulation always makes sense only if you accept that the lowest possible price is always desirable, no matter the positive effects such regulation has. In reality, it's a trade-off between security and other desirable social ends and price and the level of investment. Whether that tradeoff is worth it (and to what extent it is worth it) is a matter of debate, not an eternal truth.
Another illustration is the Pareto optimum. It is the closest economics comes to a truly objective optimum. What it means is that the productive capacity of the economy is fully utilised.
However, if someone advocates against change simply because "we are near the Pareto optimum," he is either ill-educated or has an agenda. The Pareto optimum is not a point, it is a function. All combinations of the production of goods and services which fully utilise the economy's productive capacity satisfy the Pareto optimum. The question of which combination should be chosen is a matter of preference. More healthcare or more real estate? More guns or more wind plants?
3. In conclusion
To conclude, economics and other social sciences make use of models, conclusions, studies, observations, experiments, discourse and other means by which they analyse social phenomena and make predictions. Their tools may be more or less sophisticated, and closer or further from reality.
Here we have the first pitfall of the social sciences: Pride. Every epistemology has its limits, and every model has its flaws. When a social scientist ignores these truths, he runs the risk of error through pride. "My models are perfect!" Certainly, until social circumstances change or a wild variable enters play.
The second pitfall of social sciences is mixing the positive and the normative. We can analyse what is, we can analyse actions, their possible effects, advantages and disadvantages, but the choiice of action, the choice of policy is a matter of preference, not an eternal truth ingrained within a fact. Presenting a possible solution as inevitability, consciously or not, is a failing of the social sciences, although, of course, everyone has his or her opinion on what ought, and of course any social scientist will offer recommendations.
This brings us to a problem which has been hinted at, but which constantly dogs social sciences: Politics. I know of many studies which propose solutions which would create a society that would be far better, according to the views of the majority of the population - or propose actions regarding ignored problems.
But they are ignored.
The reason is simple. In the natural sciences, any direct interest is practically non-existent. Aside from egoes and academic reputation, nothing is at stake regarding the question of quantum spin (or whatever) of some particles. When a study shows that political parties (and the general populace) don't know shit from shinola regarding how to solve the crime problem, then the study runs contrary to innumerable vested interests. What will a politician do? Say, "sorry, guys, I screwed up, let's do what the study suggests?" Sure, it may happen, but the chances are precisely the same as my being hit by a meteorite. And so such studies will languish in forgotten drawers.
Conversely, social scientists who, consciously or not, in good faith or not, accurately or not, ideologically or not, reinforce popular ideologies and prejudices will be heard, honoured, and deferred to. The left will not listen to Telser, even when he is right - he is, after all, a neoliberal. The right will not listen to Krugman, even when he is right. He is, after all, a Keynesian. And anyone who dares suggest that our strategy with regard to paedophiles might be counterproductive will be publicly crucified, although probably not literally, no matter his arguments.
In short, social sciences are enormously useful, as long as they know their limitations and stay true to themselves. When they don't, disasters happen.
A. Kaufmann, Uvod v filozofijo prava (Introduction to the Philosophy of Law),
Goode and Ben-Yehuda, Moral Panics: A Social Construction of Deviance (2009),
L.G. Telser, A Theory of Self-Enforcing Agreements (1980)
Erich Schanze, Notes on Models of Choice, Incomplete Contracting, and the Agency Framework (1990),
Erich Schanze, Economism, Legalism, and Professional Attitudes Toward Institutional Design (1993)
Roberto Saviano, Gomorrah