Is Economics a science? If so, in whIs there a rational basis for making large-scale (world/country) economic decisions?
May 31, 2008 7:29 AM   Subscribe

Is it possible to rationally settle on an economic theary?

I'm pretty ignorant about (world) economics, though I've read a few books. It strikes me that there are two major economics theories in play in the US. Grossly oversimplifying, they are (1) spread the wealth and (2) make the rich richer, and the wealth will trickle down, enhancing society as a whole.

I'm not interested (in this thread) in debating which one is right or even whether I'm off base about the theories. If I'm wrong about the nature of the theories, it's enough that there are competing theories of some sort.


Let's say you have no emotional baggage connected with either theory. You just want to back whichever one is most likely to benefit society. Is there ANY rational way to pick without "just going with your gut" or "going with the way you were raised"?

Aren't economies chaotic systems (as in Chaos Theory)? I don't see how you can reason from history or from "experiments" in other countries. Those are massively dirty test tubes. The US in 2008 is not Russia in 1950 or even the US in 1980.

The only workable test I can think of is to, say, take the U.S. right now, try one system, rewind time to the exact same point, and then try another system. Even then, you'd only learn what would work/not-work in one country, at one point-in-time.

Yet people have very strong attachments to specific economic theories. It's not strange for people to be irrationally attached (it happens to me just like everyone else). So it's tempting for me to conclude, "Oh, there is no rational basis for choosing one theory over another. It's just one of those things, like religion or whatever. It's just about strong feelings and hunches."

But I don't want to assume that's true. I know there are well-paid, well-educated people called Economists. They suggest -- presumably based on evidence -- that we should do this or that.

Is Economics (in any sense) a science?
posted by grumblebee to Society & Culture (54 answers total) 21 users marked this as a favorite
 
Great question. I know there are MeFites with more knowledge and experience in the field than I, but I have taken a couple introductory-level economics courses, and I came away from it having decided that Economics was not a science. Whatever kind of metrics are used (GDP, Unemployment, CPI, etc) are oversimplifications at best, and there seems to be little to no predictive value to them or to the systems that combine and compare and contrast them, at least on a scale as large as a national economy. To me, politics seems to play as large a role in economics as science, and that is massively frustrating.
posted by Rock Steady at 7:41 AM on May 31, 2008


Traditional capitalism is based on unlimited natural resources and continual growth of the economy. Take that as you will.
posted by sonic meat machine at 7:42 AM on May 31, 2008


Economics at it's best is a science, but as you point out there are severe data problems for certain questions. The data problems lead to contradictory results. Every science makes simplifying assumptions, some areas of economics make more of them.

At some level any real world system is going to be chaotic and unpredictable but the things you seem interested in (GDP and the Lorenz curve) can be usefully predicted. I think it is really the trade off between the two that is poorly understood, leading to debate over the best policy options. I consider most of economics a science, but not as good a science as I'd like.
posted by thrako at 7:47 AM on May 31, 2008


Years of observing our Dear Leaders has led me to conclude that economics as used to justify public policy is pretty much voodoo (or maybe a cargo cult, with the Undistorted Market taking the place of John Frum).

There do seem to be some occasional bits of genuine science going on, but economies as a whole are still too complex to nail down precisely. The general principle does seem to be, as you observe, that there's no general principle.
posted by flabdablet at 8:03 AM on May 31, 2008


Good question. Economics is a game that we play with our lives. It's a constant trade off in human values when faced with scarcity. The theories that serve economics are also political and philosophical, often absolute, because the game is dangerously complex and includes power, corruption, and internal and external threats.
posted by Brian B. at 8:05 AM on May 31, 2008


Two fields of economics that try to adhere closely to scientific principles are Econometrics (application of statistical analysis to economic theories and data) and Experimental Economics (close to behavioural psychology). Their success is open to debate, but the approach at least fits the bill.
posted by Bodd at 8:10 AM on May 31, 2008


flabdablet brings up a good point, that economics as discussed by politicians and pundits is a far cry from academic economics.

There do seem to be some occasional bits of genuine science going on,
I don't understand, didn't the guys in that article get it wrong?
posted by thrako at 8:13 AM on May 31, 2008


Economics is a social science. Very rarely is it possible to do contained experiments in the way that microbiologists can. However, even microbiologists don't know everything, nor are they able to predict with ultimate certainty what the results of their experiments will be (or why do them?). So why should we expect economists to have all the answers? As you note, society is constantly changing, but that doesn't make it completely chaotic. For example, we can trace the history of exploration through archaelogical evidence and written accounts. Humans spread from Africa to cover the entire earth in a fairly logical fashion - not chaotic at all when you consider the environment in which each group of people lived. It's not "just a hunch" to look at archaelogical evidence and conclude that one group probably invaded another based on pottery fragments. It's also not "just a hunch" to look at economic trends and make assessments or predictions. Sure, the predictions can turn out to be wrong. How many times have we been told we're close to a cure for AIDS or cancer? Yet that doesn't invalidate biology as a science.
posted by desjardins at 8:20 AM on May 31, 2008 [1 favorite]


Response by poster: How many times have we been told we're close to a cure for AIDS or cancer? Yet that doesn't invalidate biology as a science.

Yes, but those aren't scientific predictions. Those are generally predictions made by people in the news media. A scientific prediction can never be "we're close to discovering X." A scientific prediction states that natural system A will behave in way B.
posted by grumblebee at 8:33 AM on May 31, 2008


Response by poster: I consider most of economics a science, but not as good a science as I'd like.

Isn't science (a.k.a a field that uses the Scientific Method) binary? It's science or it's not, right? What is "not so good" science.

If I test a drug on 10,000 people but don't control for the placebo effect, that's not so-so science. It's not science at all. It's not using the Scientific Method, which insists on (amongst other things) controls.

I guess my other question is -- if large-scale Economics isn't science -- why should we use it? If the answer is "because it's the best thing we have," does that mean it's better than chance at making predictions? Can be verify that?
posted by grumblebee at 8:38 AM on May 31, 2008


Best answer: This is a question of the philosophy of science. If you want to decide whether economics is science, you need to define both of those first.

In all of science, one of the big questions is the problem of induction. How do we justify induction, something we do reflexively? Hume said that basically, we can't. Most people are happy to continue along with induction, because... well, what the hell else are you going to do?

A guy named Karl Popper tried to resolve this problem in the context of science (where it's influence is particularly acute) by pushing the centrality of falsifiability. These days, not many people buy this "solution" to the "problem", but falsifiability is still important. Falsifiability is necessary, but not sufficient (some, of course, disagree). I mention all this because I think one of the stronger criticisms of economics as science is that it's theories don't seem falsifiable.

If we go through the list of "basic concepts" taught to economics students, we see a lot of very sensible mathematics. It seems (though I'd be happy to hear otherwise) that these concepts are taken as axiomatic (not backed up by any data). As hypothetical, they make sense to me, but I'm skeptical that they map on to real world situations in any meaningful way. How would you even test if that were the case? As you said... those are massively dirty test tubes.

You might find this article interesting.
posted by phrontist at 8:42 AM on May 31, 2008 [4 favorites]


If the answer is "because it's the best thing we have," does that mean it's better than chance at making predictions? Can we verify that?

Ha, this would be interesting... try to get some economists to commit to making solid predictions and seeing how often they turn up right.
posted by phrontist at 8:44 AM on May 31, 2008


Isn't science (a.k.a a field that uses the Scientific Method) binary? It's science or it's not, right? What is "not so good" science.

As phrontist points out, there is a whole field devoted to defining and studying science, so I wouldn't agree that "being science" is clearly a binary state. But I only meant to say that economics does not have all the clean cut answers that I'd like, and also that some of what gets called economics is not science by any reasonable standard.


If the answer is "because it's the best thing we have," does that mean it's better than chance at making predictions? Can we verify that?

Ha, this would be interesting... try to get some economists to commit to making solid predictions and seeing how often they turn up right.


This is exactly what academic economists (not the talking heads on TV) do. For example, you write down your model and estimate the parameters on a given data sample. Then you use the estimated model to make predictions about a separate data set. If the out-of-sample forecasts are better than chance, or better than some alternative model, then that speaks in favor of your model. This relies on the two samples being comparable, which depending on the application may be a problem.
posted by thrako at 9:02 AM on May 31, 2008


Response by poster: Most people are happy to continue along with induction, because... well, what the hell else are you going to do?

Well, there is something else you can do: flip a coin.

I mean, if induction seems to make better predictions than some sort of random picking, you should obviously use induction. If not, why bother?

I'm being Mr. Hyper Rational here, just for the sake of argument. I realize that even if induction doesn't help solve a problem, it may make people feel like they have some control over their lives.
posted by grumblebee at 9:16 AM on May 31, 2008


I mean, if induction seems to make better predictions than some sort of random picking

Ha! How could you establish that induction makes better predictions? Only by induction!
posted by phrontist at 9:41 AM on May 31, 2008


Response by poster: If that's true, phrontist, is there any rational reason to use induction as opposed to flipping a coin or rolling dice?
posted by grumblebee at 9:45 AM on May 31, 2008


Then you use the estimated model to make predictions about a separate data set. If the out-of-sample forecasts are better than chance, or better than some alternative model, then that speaks in favor of your model. This relies on the two samples being comparable, which depending on the application may be a problem.

The problem here is that it's really easy to cherry pick data sets to make things work. When your theory doesn't work for a data set, you throw it out. I have very little experience with academic economists (though living in DC, I've met a lot of the NGO/Worldbank/IMF kind) but it seems like this would be a hard thing to avoid.

Regardless, finding some decent statistical rules of thumb isn't science. It's not even explanatory... it's just a lot of mucking with numbers. If the only criteria is making predictions, you could set up some genetic algorithms to pore through financial data and breed through random chance functions that are predictive. In the end, you wouldn't have any better an idea of how the underlying system works, and because of that, you wouldn't know when to expect the rule to break down.
posted by phrontist at 9:46 AM on May 31, 2008


The problem with economics is that's like a hybrid of social science and math. It's not a "real science" like biology or physics. A lot of the predictions it tries to make can never really be experimentally validated.

My feeling is that there is a lot of interesting math being done, lots of interesting research being done, plenty of good ideas being generated, but ultimately economics as a whole is beset by people making predictions which are fundamentally unsound.
posted by delmoi at 9:55 AM on May 31, 2008 [1 favorite]


Response by poster: Another thing that strikes me as weird: I never see this question discussed anywhere. I mean, I'm sure it's discussed in academic circles, but I never see it in pop culture. From responses here, it sounds like (whether it's right or wrong), there is a strong notion amongst some people that Economics is junk science.

Even though most Americans believe in God, one still hears atheists airing their points of view. One hears skeptical voices discussing psychics, UFOs and bigfoot. Yet when someone espouses an economic theory, what I mostly hear as a counter is, "That theory is wrong. Here's the RIGHT theory..."
posted by grumblebee at 9:55 AM on May 31, 2008


If that's true, phrontist, is there any rational reason to use induction as opposed to flipping a coin or rolling dice?

According to Hume, no. The way I see it, it's something almost all of us take on faith (some set of those considered mentally ill or handicapped probably don't see the world in inductive terms, and the theoretically enlightened buddhist monk may see their reliance on induction for what it is moment by moment, and maybe someone doing a lot of acid, but these cases would be quite exceptional).

Stanford's Encyclopedia of Philosophy gives greater background... needless to say a lot of philosophers have thought about this since.

If you're interested in this, I'd recommend Samir Okasha's Very Short Introduction to the Philosophy of Science, which is part of an excellent series by Oxford University Press. I'm an engineer, not a philosophy major by any means, and it was very readable.
posted by phrontist at 9:56 AM on May 31, 2008


Response by poster: Thanks. I've been hoping, for a long time, that a huge box would arrive on my doorstep, I'd open it and find the entire "Very Short Introduction" series inside. Hasn't happened yet.
posted by grumblebee at 10:02 AM on May 31, 2008


From responses here, it sounds like (whether it's right or wrong), there is a strong notion amongst some people that Economics is junk science.

Well, it's hella arcane. I'm somewhere above a mathematical layman, read The Economist every week, and it's still incredibly dense to me. I can't bring myself to take a class in it because I just find it so hard to swallow the things you'd cover on day one.

Hillary Clinton is a skeptic when it's politically expedient, but she was torn a new one by the pundits for it.
posted by phrontist at 10:07 AM on May 31, 2008


Aren't economies chaotic systems (as in Chaos Theory)? I don't see how you can reason from history or from "experiments" in other countries. Those are massively dirty test tubes. The US in 2008 is not Russia in 1950 or even the US in 1980....

If I test a drug on 10,000 people but don't control for the placebo effect, that's not so-so science. It's not science at all. It's not using the Scientific Method, which insists on (amongst other things) controls.

The double-blind true experiment with a control and treatment group is only one of the multiple methods used by scientists. Many phenomena (clades, the precession of Mercury, and earthquakes to name three) can only be approximated within the scope of a double-blind true experimental study. So for many fields, pseudo-experiments, correlational studies, case studies, longitudinal, and even qualitative studies are considered to be valid ways of building knowledge, with inferences built on triangulation of multiple lines of evidence. Darwin's Origin of Species is a great example of using multiple qualitative lines of research to build a solid theory with predictions.

For that matter, to counter something dropped by desjardins, some microbiologists don't do true experimental studies. You can't ethically or even physically create all the effects of a pandemic within a true experiment, and we know from surveys that there are thousands of species that have never been successfully cultured under controlled condition. And on the other hand some economists do perform true experiments to test some of the underlying assumptions of how people make economic decisions. So its not an either/or.

And yes, sometimes medical research is done without a placebo, especially when the risks of a placebo are both well-known and so dangerous that a placebo trial can't be ethically justified. In that case, the outcomes of an experimental treatment are compared to the outcomes of a well-established existing treatment for which there are thousands of data points.

If that's true, phrontist, is there any rational reason to use induction as opposed to flipping a coin or rolling dice?

Here is a thought experiment for you. If you honestly believe that induction (which is also the foundation of the true experiment) yields predictions no better than flipping a coin, I challenge you to go to a busy intersection, flip a coin, and take three steps into traffic every time it comes up "heads."

I'll go with Pragmatist philosophy on this one. While "induction" may not be an iron-clad method of creating knowledge, its predictions are usually good enough for doing useful work in the universe.

delmoi: The problem with economics is that's like a hybrid of social science and math. It's not a "real science" like biology or physics. A lot of the predictions it tries to make can never really be experimentally validated.

Neither can many predictions made in biology or physics. If you are going to hold this high standard, then you are forced to reject General Relativity, Evolution, and just about every theory that suggests the universe was not created last Wednesday.

And if you accept the same kinds of evidence used to support the theories of General Relativity, Evolution, and the theory that the Universe was created several billion Wednesdays ago, then you should be willing to accept those methods as used to talk about Human Behavior.
posted by KirkJobSluder at 10:09 AM on May 31, 2008


You need to distinguish economics from ideology.

While there many different ideologies which apply and misapply economics, there is surprisingly little partisanship among economists when they do economics. Agreement tends to be widespread on the core mechanics of the creation and transmission of wealth, while controversy tends to focus on theories which underly the mechanics and epiphenomena around them. Most importantly, you will never predict what a PhD economist has to say about many economic questions of interest by knowing whether he's a Republican or a Democrat. (I suspect that you will find more Democrats than Republicans who will qualify that by saying, when it comes to forming public policy, that there are things more important than economics, and they will thus tolerate a degree of redistribution and regulation that they might concede isn't maximally efficient.)

You can see the politics - economics divide in very bright relief if you habitually read Paul Krugman's column. He's predictable and conventional when he talks politics, but he gets excited and interesting, and surprisingly free of East Coast liberal party line (such as his criticisms of Obama) when he actually starts to talk about matters where economic theory matters.
posted by MattD at 10:16 AM on May 31, 2008 [1 favorite]


phrontist: Regardless, finding some decent statistical rules of thumb isn't science. It's not even explanatory... it's just a lot of mucking with numbers. If the only criteria is making predictions, you could set up some genetic algorithms to pore through financial data and breed through random chance functions that are predictive. In the end, you wouldn't have any better an idea of how the underlying system works, and because of that, you wouldn't know when to expect the rule to break down.

Science is a collaborative and community process that repeatedly tests theories against multiple sets of data.

How do we test good explanations of mechanism? By making predictions that are compared against additional data.

And in some cases, a good descriptive theory beats bad explanatory theory. Ptolemy, Copernicus, Tycho, and Kepler had beautiful explanatory theories that were fundamentally wrong. It was Kepler's entirely descriptive theory of the motion of planets that paved the way for modern astronomy.
posted by KirkJobSluder at 10:28 AM on May 31, 2008


Phrontist is right, and judging by your allocation of a best answer, I suspect that he addressed this substance of your question. But you asked a lot of questions, and I'd like to come at this from a different angle, with a few points that hope to get at the meat of your problem.

1) It seems that your first assumption of two different "theories" is probably a bad starting place. I don't even know where there's a "line" between positive and normative economics, but it's a distinction you should keep in mind as you attempt to answer these questions.

2) Your question is about Economics writ large, but your context targets Macroeconomics. Taking a step back, though, let's start small and work bigger.

2a) I can build a model to predict the price of a liquid corporate bond, given the term structure of treasury interest rates and a benchmark credit spread. This pricing model, as they say, is nothing more than a "useful lie," but it's surely more accurate than the output of a random number generator.

2b) Scaling this up, I can also build a model to predict the level of a stock index, given trailing earnings and interest rates. This model will not be as accurate as the model in 2a (I don't suggest using it to try to make money) , but again it will be better than a number that one simply plucks from the air.

2c) Scaling up once more, one can use Fairmodel to forecast GDP. There's no way, of course, that the model can predict the actions of every agent in every economy, and of course its outcome will be wrong, in some sense. But it's once again better than a randomly generated number. The model is falsified every time, but it is nonetheless usually a useful predictor/descriptor of reality.

3) Economics isn't really a chaotic system, in my view, but more of a complex adaptive system.

4) Going back to 1), and to your "two theories", it's not even clear what the arguing factions are debating. The way that political economy is practiced in the public arena, few even bother to identify what it is that they're trying to maximize. Even if one begins at pure utilitarianism, well, what's the relevant population? Americans? The world? People alive today? People not yet born? Does fairness matter? et cetera.
posted by Kwantsar at 10:33 AM on May 31, 2008 [2 favorites]


Response by poster: Here is a thought experiment for you. If you honestly believe that induction (which is also the foundation of the true experiment) yields predictions no better than flipping a coin, I challenge you to go to a busy intersection, flip a coin, and take three steps into traffic every time it comes up "heads."

There's a difference, though. Most intersections are like each other, at least on the scale that's important to your thought experiment. So I DO tend to trust induction in cases like that (because it seems to work for me).

But let's expand your thought experiment: each intersection is unique, even at the general, human-interaction-with-intersections scales. At some, don't-walk means walks; at some, planes land instead of cars; at some, bombs go off regularly; etc. At this point, I stop trusting induction. And flipping a coin really does seem as good as any other means. (And if I had to cross the street, I might just go ahead and flip one).

My claim is that large-scale economies may be like those sorts of intersections. I don't get how they're close enough to each other to even suggest that induction (if it works in any circumstance) will likely work with them.
posted by grumblebee at 10:39 AM on May 31, 2008


Response by poster: Kwanstar, I'm trying to assess if the popular understanding of large-scale (country level) economics is in any way rational.

I'm assuming that all participants want general economic prosperity, meaning they would like -- say -- all people (or as many people as possible) to have enough money to afford, clothes, shelter, food, etc. beyond the bare subsistence level. In other words: no more poverty or greatly reduced poverty. To keep things relatively simple, lets say I'm talking on the country level, not on the planet level.

I realized that plenty of people don't care about this. Some people just care about themselves, etc.

But I have heard plenty of people, who genuinely seem to want to help the majority -- not just, say, the rich. Some of these people seem SURE that the way to do this is to spread the wealth; others seem SURE that the best way to do this is to feed the top and watch the wealth trickle down.

My question: is it rational to hold either view. Or is the rational view, "We just don't know."
posted by grumblebee at 10:50 AM on May 31, 2008


Here is a thought experiment for you. If you honestly believe that induction (which is also the foundation of the true experiment) yields predictions no better than flipping a coin, I challenge you to go to a busy intersection, flip a coin, and take three steps into traffic every time it comes up "heads."

I brought up induction only in passing, to arrive at the greater point of falsifiability. Science is the best way to carry out inductive reasoning. Inductive reasoning is the basis of almost everything we do, and I'm not arguing against it, just pointing out an interesting philosophical conundrum.

Anyway...

And in some cases, a good descriptive theory beats bad explanatory theory. Ptolemy, Copernicus, Tycho, and Kepler had beautiful explanatory theories that were fundamentally wrong. It was Kepler's entirely descriptive theory of the motion of planets that paved the way for modern astronomy.

Lets say there are predictive economic models. That still doesn't mean they're descriptive. Collecting lots of data and finding correlations does not, on it's own, produce scientific conclusions. You need to justify the relevance of what you're measuring and myriad other assumptions that qualify your results. Hence catchphrases like "correlation is not causation".
posted by phrontist at 11:23 AM on May 31, 2008


I think it's quite rational to say we don't know much about how economies behave at a large scale, compared to how, say, electrons behave in a conductor.
posted by phrontist at 11:28 AM on May 31, 2008


But I have heard plenty of people, who genuinely seem to want to help the majority -- not just, say, the rich. Some of these people seem SURE that the way to do this is to spread the wealth; others seem SURE that the best way to do this is to feed the top and watch the wealth trickle down.

The latter certainly hasn't shown itself to work at any time in the past, the former has. No scientific conclusion can be made either way.
posted by phrontist at 11:29 AM on May 31, 2008


grumblebee: My claim is that large-scale economies may be like those sorts of intersections. I don't get how they're close enough to each other to even suggest that induction (if it works in any circumstance) will likely work with them.

Even with my casual understanding of economics I know that not all economists are that concerned with focusing on large-scale economies, and I know that there is some purely descriptive work going on in economics, so I'm not convinced that your criticism of the field as a whole is accurate.

To use your thought experiment, while the details of an intersection might change, there are likely some common situations on which I could make a judgment better than chance. If I go to Japan or England, everyone might drive on a different side of the road, and perhaps use different signals, but I know that the positive and negative acceleration of vehicles are going to be constrained by some basic physics and human factors. A 30km/s^2 acceleration is probably unlikely. I have a lifetime of experience gaging distance and speed, and if I need to, I can observe for 20 minutes and make some inferences based on that behavior.

So, an economist would open up that precious snowflake of an economy and take a careful look at what's inside. And their predictions are probably going to be hedged with a lot of, "well, the economy here is like England in some respects, and like Bulgaria in some respects, but this model does better than chance in looking at historic data for the last 20 years, therefore it's worth publishing and hopefully it will do better than chance with data for the next 5 years."
posted by KirkJobSluder at 11:33 AM on May 31, 2008


I don't understand, didn't the guys in that article get it wrong?

Their model made incorrect predictions. What makes it science is that they tested their model's predictions against the real world, rather than simply demonstrating that it was consistent with some axiom set or other.
posted by flabdablet at 11:33 AM on May 31, 2008


Lets say there are predictive economic models. That still doesn't mean they're descriptive. Collecting lots of data and finding correlations does not, on it's own, produce scientific conclusions. You need to justify the relevance of what you're measuring and myriad other assumptions that qualify your results. Hence catchphrases like "correlation is not causation".

Two words: Hubble's Law. A correlation is an entirely valid scientific conclusion and provocative correlations, when published, are often the stepping stones to building a great theory.

"correlation is not causation" is an overused catchphrase. The fact of the matter is that almost all of the theories we have regarding real-world phenomena are built around multiple lines of correlational data.
posted by KirkJobSluder at 11:41 AM on May 31, 2008


A mathematician, an accountant, and an economist are all waiting to be interviewed for the same job. The interviewer calls in the mathematician first and asks "What do two plus four equal?" The mathematician answers "Six." The interviewer says "exactly six, every time?" and the mathematician gives him a confused look and says "Yes, exactly six every time."

Next, the interviewer calls in the accountant and asks the same question -- "What do two plus four equal?" The accountant answers "Six, on average. Plus or minus ten percent, but six on average."

Last, the interviewer brings in the economist and asks the question again: "What do two plus four equal?"

The economist glances around the room, locks the door, pulls down the windowshade, dims the lights, and whispers to the interviewer "What do you want it to equal?"

Many economists are aware of the futility of trying to mathematically model a system as complex as an economy -- as the joke above suggests, it's a bit of a running gag. But attempts to make economics fit the "pure science" mold, influenced by the models of Newtonian physics, have resulted in today's emphasis on simplified models that just don't always work. There are schools of economic thought that reject this emphasis -- most notably the Austrians, who say that human action is too complicated to model, and instead try to just deductively reason economics from a couple basic axioms.

I think one large component of popular confusion about economics comes from failure to draw a distinction between positive economics, which asks "hey, what's going on in this economy and how can we understand it?," and normative economics, which asks "what should be going on in this economy to achieve the things that I want?" Milton Friedman wrote a famous essay on this distinction (and the question of the relevance of modeling) that's worth a read.

When you say that economists "suggest -- presumably based on evidence -- that we should do this or that," you're getting tangled in the positive-normative distinction. Same thing when you start with an idea that the goal of economic theories is "enhancing society." Sure, policymakers use economics for whatever ends they hope to promote (often, "enhancing society"), and many economists study the effects of public policy. But a good economist doesn't ask "is this policy a good one," but rather "will this policy have its intended effect?"

Of course, none of this settles the question whether economics is a "science," but it's an important consideration.
posted by ecmendenhall at 1:46 PM on May 31, 2008


The problem here is that it's really easy to cherry pick data sets to make things work. When your theory doesn't work for a data set, you throw it out. I have very little experience with academic economists (though living in DC, I've met a lot of the NGO/Worldbank/IMF kind) but it seems like this would be a hard thing to avoid.

This is a general problem for any field that uses statistical models (not just economics) and there's all sorts of work on understanding this and avoiding researcher bias in evaluating such models. I don't know about economics, but in machine learning, cherry picking data sets would prevent your research from having a chance of getting past peer review, and it would be extremely difficult to hide that you've done so. The point is that without some more specific criticism (i.e. about the ways researchers in a particular discipline try to avoid this bias), this isn't a real problem for anyone, it's just a problem that might in principle exist for very naive versions of some field of study.
posted by advil at 1:46 PM on May 31, 2008


One problem is defining what one means by "benefiting society". Does it mean full employment, or protecting the environment? Some of each? Well, how much exactly; what's the trade-off? If it's making people happier, then perhaps one might want to emulate Denmark (the supposed happiest nation) or Bhutan's concept of Gross National Happiness. But since the rich and powerful have a greater hand in the economy than the rest of us, perhaps their happiness is allocated a greater degree importance.

Even if a set of goals and a corresponding economic policies is figured out, there's nothing to stop some group of people from "gaming the system" to profit themselves at the expense of the common good. Examples here might include the current mortgage crisis.

Sorry if this doesn't answer the question. But I think reasons like this are part of why people vary so much on economic theories. Of course, the "emotional baggage" component is also huge.
posted by DarkForest at 2:19 PM on May 31, 2008


advil is right, disregard that comment.

stepping stones to building a great theory

I don't deny that finding correlations is important in developing theories, but I think economists aren't finishing the process. You still need to come up with some theory for how the correlation came about, and demarcate when it should work. The theory should then make quantitative predictions that you can test. This is how you tie your qualitative theory to your quantitative data.

Statistical significance, used to "grade" correlation, does not mean anything on it's own. If I'm doing an experiment to verify Ohm's law, and I measure a given amount of current through a given resistance and it doesn't fit the theory... there are a few possible explanations. The first, and most important, is that ohm's law does not apply to the situation... it is wrong. This is a very important possibility, and it's what makes science legit. Another possibility is that my measuring apparatus is uncertain to some degree. Is the measured value within my margin of error? Ah, it is, so the theory still holds (I should keep making more precise measurements to be sure, but you never prove anything in science... only advance better theories).

So in economics, we have the generation of qualitative theories and something called "econometrics" which is supposed to provide quantitative verification. So lets we have a theory, "longer skirts and other details of women's dress infulence economic prosperity" (this is an actual correlation that actually held very well for many years). We collect data about skirt length and GDP for many countries around the world and find, to our great joy, that there is a very strong correlation... it is statistically significant.

The big problem here is that statistical methods, regardless of their sophistication, are about identifying signal to noise. We're trying to analyze how many small measurement errors influence the final result. The assumption here is that if your theory is right, it's predictions perfectly match the data. Should their be some deviation, either your theory fails or the deviation is within a range explicable by measurement error.

Economists, as far as I can tell, don't work like this, as other (real?) scientists do (like astronomers in your example). They come up with a qualitative theory ("skirt length and GDP are related") and then they collect a bunch of data to see if they find a correaltion. Through sophisticated and impressive number crunching they find some numbers that relate the various variables they've established to one other. If it doesn't work perfectly in all cases, it's still a good "theory" because the correlation is pretty strong. How strong does it need to confirm the theory? How about a correlation coefficient of 5%. Huh? Well, maybe not always 5%, but we'll just pick a number that's "good enough".

This isn't science in the same way astronomy, physics, biology, chemistry, or (well conducted) medical science is. The quantitative results don't have rigorous logical connections to the qualitative theories they ostensibly set out to prove.

This academic, with econ credentials, rants about this at length (I only glanced over it, borrowed her skirt example). I've seen better critiques of the discipline, but I can't find them right now... it's driving me nuts.
posted by phrontist at 2:41 PM on May 31, 2008


Let's say you have no emotional baggage connected with either theory. You just want to back whichever one is most likely to benefit society.

Let's cut to the chase. "Most likely to benefit society." First, you have to narrowly define "most likely," "benefit" and "society."

I don't mean to sound glib, but there's wide interpretations for all three terms.

Most likely -- Is this everyone? Or a majority?
Benefit -- Benefit how, exactly? Money in your pocket? Health care? Personal freedom?
Society -- Who is in or out of the society? Everyone in the country? The planet? Shareholders of corporations? Citizens only? Illegal aliens?

You can't "rationally determine" anything until you first define your parameters.

And here, IMO, is where most people fall down when comparing and contrasting economic systems -- they simply fail to identify what it is exactly they would like to see happen.
posted by Cool Papa Bell at 2:51 PM on May 31, 2008


I should add that I still the economists do very useful things, like set the interest rate and make predictions. We all act in an unscientific manner pretty much all the time - it would be almost impossible to do otherwise. There is a difference though between rational guessing and true science.

I can be as certain about gravitation or ohms law (with certain caveats about precision and circumstance) as I can about anything because of it's very good scientific backing. There is a very long list of carefully considered assumptions and empirical experiments backing all of them up save for a few axioms (for instance... the at least 3-dimensional Cartesian nature) of space. Lets say new experiments return values that can't be explained by experimental or measurement error. It means our theory is wrong, we made a mistake in one of our assumptions - back to the drawing board.

Contrast this with an economic "theory". Lets say our coefficient of correlation starts getting more loose as new data comes in. At what point do we throw it out? 6%? Nah, relax man, it still works pretty well...
posted by phrontist at 2:54 PM on May 31, 2008


I should add that I still the economists = I should add that I still think economists
posted by phrontist at 2:55 PM on May 31, 2008


To be totally, absurdly clear: choosing an arbitrary acceptable coefficient of correlation is a philosophical problem because it make the determination of falsification subjective.
posted by phrontist at 3:01 PM on May 31, 2008


I think it's a scope issue- all science began as "hey, look at that! I wonder what makes that happen?" And so people experiment and learn about the universe.

But economics is so huge and the variables are so vast, that it is practically unknowable. But that doesn't mean that truths don't exist- it just mean that it's really hard or impossible to know what they really are.

It's also compounded by the fact that the economy is constantly moving and circular. My spending is your income is your spending is my income. Forever. Or until the robot holocaust.

But there are certain things that we've learned that are absolutely true- if, magically, tomorrow, every dollar everyone in the world had were decreed to now be a million dollars, nothing would change. Everyone would still own exactly the same share of the pie. Now that is one big, complex, ever expanding and contracting pie, but the fact remains that there IS a finite amount of value "out there" at any given time. And the value of anything is ONLY what you can use it for, or what someone will give you in return at the moment you want to sell it. That's the essence of economics. Everything else is masturbation.

(An example of not understanding the scope of things, and the danger of generalizing things that cannot be generalized. The Laffer Curve. The theory is as basic as it can get- at 0% taxation, the government cannot do what it needs to do. At 100% taxation, nobody will have anything left with which to spend because the government will have taken it all. And that somewhere in between, an optimum level of taxation exists where the government generates the needed amount of revenue with the least amount of burden on the economy. And, further, that if you are on the high side of that optimum level, reducing the level of taxation can increase revenue. If the governmental version of the supply versus demand curve- at that point where the lines intersect everyone is extracting the greatest value from their resources. But dunderheads confuse and misunderstand it, and all of a sudden you have clowns saying that all tax cuts pay for themselves- a blatant bastardization of the concept.)
posted by gjc at 4:25 PM on May 31, 2008


But there are certain things that we've learned that are absolutely true- if, magically, tomorrow, every dollar everyone in the world had were decreed to now be a million dollars, nothing would change. Everyone would still own exactly the same share of the pie. Now that is one big, complex, ever expanding and contracting pie, but the fact remains that there IS a finite amount of value "out there" at any given time.

Even this isn't true. Take for, example, stock prices. A company issues 1000 units of stock, at $5 each to be traded. Various factors cause the price to rise to $10 (the public sees the comapany as a growth opportunity, it offers dividends, etc.). At some point they decide they want to make it easier for the little guy to make an investment in their company, so they annouce that all shareholders get twice the number of shares. What happens? If people were rational agents as posited by traditional economics, the share price would simply halve, right? TIme and time again it seems to drop lower than that, because people feel weird about seeing low numbers.

And the value of anything is ONLY what you can use it for, or what someone will give you in return at the moment you want to sell it.

This is a definition... worth in an economic context is defined as what you can get for it. It's not a finding.
posted by phrontist at 5:21 PM on May 31, 2008


phrontist: I think you putting science on a pedestal, and saying a lot of things that are just plain ignorant and misinformed.

To start with, your hypothetical Oms' Law exercise wasn't an experiment, it wasn't even a pseudo-experiment.

And let's just say that you do run an experiment, and you find results that you think justify rejecting the null hypothesis, and accepting the hypothesis predicted by your theory. How do you cut through the measurement noise to make this claim? You will use a statistical test that is a mathematical special case of the correlation coefficient. And just about every criticism you raise regarding correlational studies are also criticisms of experimental designs and their results.

In fact, I'll say this as a person who has worked in biology and the social sciences, and still reads the primary literature for both, if you do not give me those statistical estimates of error and the results of the special case of the correlation coefficient, I'll start to think you are pulling one over on me.

Second, you gloss over the problem that in many cases, an theory cannot be tested using an experimental design. We can't produce the high gravitational masses described by General Relativity, but we can look at thousands of datapoints of gravitational systems ranging from white dwarfs to galaxy superclusters.

Third, NO, it is not the case that all scientific theories demand a perfect match to the data. Just as an example, geneticists have known going on 70 years now that phenotypes are complex multi-factoral systems in which you might have multiple genes and multiple environmental influences determining something like, susceptibility to breast cancer. The identification of a single factor which is strongly correlated with a given phenotype is still often useful, first, because we can't understand how multi-gene phenotypes work without teasing out those individual factors, and second because aggressive detection and prevention may be a good idea for people with those markers.

Contemporary scientists in all fields struggle with multi-factoral and complex systems, and dirty data. As with the previous cited example of Hubble's Law sometimes the interesting statement that two variables among the thousands of possible variables appear to be related are good starts.

To be totally, absurdly clear: choosing an arbitrary acceptable coefficient of correlation is a philosophical problem because it make the determination of falsification subjective.

Well yes, this is a problem that's shared by study designs that use derivations of the coefficient of correlation to evaluate their claims, including experiments, pseudo-experiments, and longitudinal studies. And it's a problem that is not unique to Economics. You are well aware that the Eddington observations that initially appeared to confirm Einstein's predictions, were later found to not be of sufficient power to support the conclusions?

I'm not qualified to evaluate methodological flaws in economics research. But this notion that economics is uniquely flawed because it uses correlational methodology, or because of issues regarding the significance of claims, or because all scientific theories make deterministic predictions about complex systems is grounded on a basic misconception of how scientific research actually works.
posted by KirkJobSluder at 5:41 PM on May 31, 2008


Or to be absurdly clear:

Correlational studies that establish which of thousands of possible factors might be significant influences in complex systems are accepted as science by peer review in dozens of disciplines. If economics is a pseduo-science it is NOT because it uses correlational methods or because it has failed to deliver deterministic predictive models.

What happens? If people were rational agents as posited by traditional economics, the share price would simply halve, right?

Of course, there are economists who test these assumptions using true experimental designs.
posted by KirkJobSluder at 6:07 PM on May 31, 2008


At some point they decide they want to make it easier for the little guy to make an investment in their company, so they annouce that all shareholders get twice the number of shares. What happens? If people were rational agents as posited by traditional economics, the share price would simply halve, right? TIme and time again it seems to drop lower than that, because people feel weird about seeing low numbers.

Ahem. It does halve in a true, traditional split. And then they all go back on sale at the halved price. And then people generally sell portions of their long-term investments in order to take advantage of a perceived, short-term, profit-taking opportunity ("I had 100 shares yesterday, and now I have 200. I'll sell 50 and make a little juice now, and I'll still be ahead of where I was before."), especially when the lower price will indeed spur stock purchases ... which usually comes after this profit-taking dip. Go look at Microsoft's performance if you don't believe me.

So, yes, it's largely rational. Not totally rational, though (hello, Pets.com).
posted by Cool Papa Bell at 8:32 PM on May 31, 2008


Yes, economics is a science, inasmuch as a 'science' is a system of hypotheses, experiments, and resultant conclusions.

The problem, and the reason it earns the sobriquet "the dismal science", is that you can't deliberately set up an experiment, they way you can in, say, chemistry.
posted by pompomtom at 12:50 AM on June 1, 2008


Paul Krugman has an interesting essay which discusses the methodology used in economics, and its limitations: The Fall and Rise of Development Economics.
... the problems of economics and of social science in general are part of a broader methodological problem that afflicts many fields: how to deal with complex systems.

It is in a way unfortunate that for many of us the image of a successful field of scientific endeavor is basic physics. The objective of the most basic physics is a complete description of what happens. In principle and apparently in practice, quantum mechanics gives a complete account of what goes on inside, say, a hydrogen atom. But most things we want to analyze, even in physical science, cannot be dealt with at that level of completeness. The only exact model of the global weather system is that system itself. Any model of that system is therefore to some degree a falsification: it leaves out some (many) aspects of reality.

How, then, does the meteorological researcher decide what to put into his model? And how does he decide whether his model is a good one? The answer to the first question is that the choice of model represents a mixture of judgement and compromise. The model must be something you know how to make -- that is, you are constrained by your modeling techniques. And the model must be something you can construct given your resources -- time, money, and patience are not unlimited. There may be a wide variety of models possible given those constraints; which one or ones you choose actually to build depends on educated guessing.

And how do you know that the model is good? It will never be right in the way that quantum electrodynamics is right. At a certain point you may be good enough at predicting that your results can be put to repeated practical use, like the giant weather-forecasting models that run on today's supercomputers; in that case predictive success can be measured in terms of dollars and cents, and the improvement of models becomes a quantifiable matter. In the early stages of a complex science, however, the criterion for a good model is more subjective: it is a good model if it succeeds in explaining or rationalizing some of what you see in the world in a way that you might not have expected.

Notice that I have not specified exactly what I mean by a model. You may think that I must mean a mathematical model, perhaps a computer simulation. And indeed that's mostly what we have to work with in economics. But a model can equally well be a physical one, and I'd like to describe briefly an example from the pre-computer era of meteorological research: Fultz's dish-pan....
Is there ANY rational way to [evaluate a proposed policy] without "just going with your gut" or "going with the way you were raised"?

My understanding is that mainstream economics is based on mathematical models which describe the main features of an economy, and that there isn't much controversy about these models, regardless of one's politics. So you can evaluate a proposed policy by running it against a model of the economic situation, and seeing what happens.

Here's an example from Krugman, analyzing policy options in Latin America in the mid-1990s.

Another example, describing French economic policy in the early 1980s:
The only thing I do remember is a conversation over dinner with an adviser to the new [Mitterand] government, who explained its plan to stimulate the economy with public spending while raising wages and maintaining a strong franc.

To the Americans present this program sounded a bit, well, inconsistent. Wouldn't it, we asked him, be a recipe for a balance of payments crisis (which duly materialized a few months later)? "That's the trouble with you Anglo-Saxon economists--you're too wrapped up in your theories. You need to adopt a historical point of view." Some of us did, in fact, know a little history. Wasn't the plan eerily reminiscent of the failed program of Leon Blum's 1936 government? "Oh no, what we are doing is completely unprecedented."
posted by russilwvong at 11:23 PM on June 1, 2008


Response by poster: My understanding is that mainstream economics is based on mathematical models which describe the main features of an economy, and that there isn't much controversy about these models, regardless of one's politics.

How successful are those models? And are there restrictions on their usefulness? If so, what are the restrictions? For instance, can we successfully model an economic system's outcome one year from now but not three years from now? Those seems like a key questions, and surely they're testable. You can test them by using plugging a real-life choice into the model and then testing the model's predictions vs. the real-live outcomes. I assume economists do this all the time. What are the results? 90% accurate? 40% accurate? 2% accurate?

If they're largely accurate, then I'd agree that politics needn't enter into it. We can just say, "Sorry, Mr. President. We ran SimEconomy and your plan won't work."
posted by grumblebee at 7:08 AM on June 2, 2008


Economics does not have predictive power, so is not a Science. Period.

But, that does not invalidate it. As a line of thought it is very useful to explain and identify patterns and correlations post-hoc, and that quality coupled with it's rationality as a line of thought makes it a million times more valid than some things that are passing as Hard Science these days (string theory, pharmaceutical trials, jack-knifing to reduce variability in stats).

I value it like I value Psychology (a whole different debate, but suffice to say I mean Bandura, not "The Secret"). Both are empirical philosophies that are valuable in developing theory, but that are not stand-alone and need Science to be proven credible or valid.

I suggest googling "The Scientific Method"
posted by doppleradar at 6:20 PM on June 2, 2008


The Laffer Curve. The theory is as basic as it can get- at 0% taxation, the government cannot do what it needs to do. At 100% taxation, nobody will have anything left with which to spend because the government will have taken it all. And that somewhere in between, an optimum level of taxation exists where the government generates the needed amount of revenue with the least amount of burden on the economy.

The Laffer curve seems to be an argument that assumes a flat tax and only a flat tax. A true progressive tax can theoretically go to 100% with none of the dangers assumed by the curve because of marginal rates because the rates increase within the tax bracket itself, not over the entire income. This is why progressive taxation is fair across the board. A 100% progressive tax would essentially be an earning cap at a given point, not a ban on income.
posted by Brian B. at 6:55 AM on June 3, 2008


grumblebee: We can just say, "Sorry, Mr. President. We ran SimEconomy and your plan won't work."

I already sent grumblebee a followup e-mail, but it's an interesting question.

To be more specific, grumblebee's question is about rationally evaluating possible policies (not theories) with respect to taxation and redistribution.

Here in Canada, at the federal level, economists in the Finance Department are responsible for calculating the effect of changes in the tax system. When introducing a new tax, how much revenue will it raise? How much impact will it have on economic activity? Or conversely, when reducing an existing tax, how much revenue will be lost? (In the US, this is the responsibility of the Treasury Department.)

How accurate are their estimates? There's an interesting discussion of the uncertainties in this annex to the 2005 federal budget: "Risks and Uncertainties in Fiscal Projections." Includes a graph comparing forecast and actual growth in nominal GDP.

Uncertainties notwithstanding, Canada's been able to run a budget surplus for more than ten years now, so I'd say that economic forecasting has a reasonable record.

Returning to the original question:
I'm assuming that all participants want general economic prosperity, meaning they would like -- say -- all people (or as many people as possible) to have enough money to afford, clothes, shelter, food, etc. beyond the bare subsistence level. In other words: no more poverty or greatly reduced poverty. To keep things relatively simple, lets say I'm talking on the country level, not on the planet level.

... Some of these people seem SURE that the way to do this is to spread the wealth; others seem SURE that the best way to do this is to feed the top and watch the wealth trickle down.
Both productivity growth (productivity = output per worker) and inequality are important.

The traditional economic view is that there's a tradeoff between growth and inequality. Erwan Quintin and Jason L. Saving (Federal Reserve Bank of Dallas):
In classical models, economic growth depends chiefly on the rate at which nations accumulate productive resources, a factor that traces to aggregate savings rates. In this context, distributional considerations matter for growth only if households’ propensity to save varies systematically with wealth. If the rich save at a high rate, a view closely associated with prominent economist Nicholas Kaldor, unequal societies can actually build up their productive infrastructure faster than equal ones, achieving higher growth rates.

Inequality could also foster growth because new industries typically require large initial investments. If credit markets function poorly, a society’s savings may not be efficiently transferred to investments. In this environment, a high concentration of wealth may allow some investors to overcome these impediments and stimulate growth by bringing capital-intensive industries into being.

In the early work, income or wealth redistribution policies are overwhelmingly viewed as detrimental to growth based on at least two arguments. First, redistribution via such instruments as progressive taxation distorts incentives to save, which reduces resource accumulation. Second, some variation in economic rewards helps provide incentives to invest and work.

The classical view long dominated economic thought and emphasized that policies designed to reduce inequality would entail adverse consequences for economic growth.
As usual with tradeoffs, the question becomes, what level of redistribution will reduce poverty to an acceptable level, while not reducing productivity growth unacceptably? It depends on what you consider to be "acceptable", which is a political rather than economic criterion. Economists can help by calculating the likely effects of a particular level of redistribution, but the actual decision is political, rather than economic.

(By the way, Quintin and Saving point out that reducing inequality can have positive effects as well as negative effects on productivity growth.)
posted by russilwvong at 11:22 PM on June 4, 2008


Economics does not have predictive power, so is not a Science. Period.

Yes it does, yes it is. Full stop.
posted by pompomtom at 8:50 PM on June 5, 2008


« Older Japan Filter   |   kung foo lite Newer »
This thread is closed to new comments.