Saturday, December 20, 2014

Noah Smith — Should theories be testable?

I don't see why we should insist that any theory be testable. After all, most of the things people are doing in math departments aren't testable, and no one complains about those, do they? I don't see why it should matter if people are doing math in a math department, a physics department, or an econ department.

I think testability starts to matter when you start thinking about applying theories to the real world. This is why I get annoyed when people ignore the evidence in business cycle theory, but not when they do it in pure theory.

Suppose you're studying the properties of repeated games. Who cares if those games represent anything that really exists today?…
 
But when you start making models that claim to be about some specific real thing (e.g. monetary policy), you're implying that you think those models should be applied. And then, it seems important to me to have some connection to real data, to tell if the theory is a good one to use, or a crappy one to use. That's testability.

Anyway, this sort of seems very college-freshman-dorm-discussion-level when I write it out like this, but I think there are a surprising number of people who don't seem to agree with it...
Exactly.

Science is not a thing but an activity. It's what scientists do. There are three major areas in doing science: 1) theoretical science (aka pure science) focusing chiefly on formalization, 2) experimental science, heavily involving design of experiments, and 3) applied science, e.g., engineering and technology, medicine, policy science, which apply general principles to specific conditions. These are separate fields, notably physics, where theoretical physics is advanced math, and experimental and applied physics are about not only formal knowledge but also practical skill. The three fields require development and use of different knowledge and skills. All are essential the scientific enterprise that results not only in expansion of the knowledge base but also in technological innovation.

What is so difficult to understand about this?

Noahpinion
Should theories be testable?
Noah Smith | Assistant Professor of Finance, Stony Brook University

10 comments:

circuit said...

I think there is a case to be made that policy research is different from other scientific activities. Policy research should be deduced from practice. This contrasts with the popular scientific hypothesis-testing approach in which social and economic phenomenon are studied primarily in order to test a theoretical framework, which is what modern macro is all about these days, unfortunately. The so-called empirical-inductive approach forces us to keep our 'eye on the ball' -- the data! This approach also tends to reduce the risk of misinterpreting the data as a result of focusing too much on a specific theory.

Matt Franko said...

Tom I think what these people are all tied up in is leading to actual changes in their cognitive abilities. .

These circuits and synapses that Roger teaches about are somehow being effected by what these people are working in.... the result is actual changes in their cognitive ability that we can see but they can't because of these cognitive effects that they manifest...

While we are left scratching our heads... and saying like you posit here 'what is so difficult to understand'?

Rsp

Tom Hickey said...

Right. It's a reason that management science went in its own direction. Harvard Business School pioneered the case approach.

Here's the coursework of the coursework at MIT's Sloan School of Management

PhD students fulfill their coursework and methodology requirements by taking advantage of the more than 150 subjects offered at MIT Sloan — in addition to hundreds more offered across MIT. Students may also avail themselves of courses at other local universities.

PhD Program curriculum at MIT Sloan is organized into three broad areas, each of which contains several research groups:

Management Science
-Information Technologies
-Marketing
-Operations Management
-System Dynamics

Behavioral & Policy Sciences
Primary and/or secondary concentrations:
-Economic Sociology
-Institute for Work & Employment Research
-Technological Innovation, Entrepreneurship & -Strategic Management
-Organization Studies
Secondary concentration only:
-Global Economics & Management

Economics, Finance & Accounting
-Finance
-Accounting

Brian Romanchuk said...

His comment about mathematics is questionable.

Modern mathematics involves taking a set of axioms (assumptions), and assumed rules of logic to generate systems of statements that are implied by those axioms. If you change the axioms, you change the "true" statements. The only way of "falsifying" a statement is demonstrating that it is contradicted by something else within the logical system.

When people apply aome parts of mathematics (differential equations, etc.) to analysing real world systems, they are not aiming to falsify the math, rather falsifying the assumed mathematical model for descrbing that behaviour.

To be honest, I am unsure whether this matters for economics. The model fits are so poor that it is hard to get excited about whether it is "scientific" or not.

Tom Hickey said...

Not sure what you mean, Brian. Math has nothing to do with reality since it is entirely syntactical, that is, based on formation and transformation rules, and the criterion in math is consistency rather than correspondence. Math models the imaginary or ideal in "logical space." Obviously, folk in science vs, pure math know that their models will have to have semantic import in order to be representational of reality, obviously a requirement for causal explanation and prediction (hypothesis formation). But theoretical scientists are not themselves overly concerned with this from what I can tell. Einstein developed a theoretical model out of his head and he was unconcerned with testing it himself. That came only later and it was a contribution of experimental science.

I have no issue with economists playing with math modeling for whatever reason. The issue is when a syntactical system, which is what a math model is, is interpreted semantically as having correspondence with events. Then the issue of testing arises, which involves predication (hypothesis formulation) in addition to causal explanation.

Some contemporary scientists apparently do not think that testability is ever an issue as long as the theory can be shown to be consistent. Other's think that causal explanation is not a necessary condition of science either but that only correlation.

Some influential conventional economists come down on the side of consistency and correlation rather than correspondence and causality. That is disputed by some other conventional economists and many heterodox economists.

It is another instance of the ongoing debate since ancient time between rationalists and realists, formalists and empiricists.

Historically, the dividing line between philosophy and science has been correspondence, which requires empirical testability rather than only consistency, and causality rather than correlation.

My question is if one rejects correspondence- empirical testability and causal explanation as the criteria for distinguishing scientific theory from speculative philosophy, what criteria do you propose? Or do you hold that there no difference between the two logically but only in relation to subject matter or methodology, e.g., degree of formalism?

Tom Hickey said...

The model fits are so poor that it is hard to get excited about whether it is "scientific" or not.

This is precisely the point. When one asserts or implies that one's work is scientific what does that mean if the result is not very representational or what is purportedly modeled?

If one wants to claim that one is just creating conceptual or math models, fine. but they don't want to do that. They want to claim that the model is useful because it is representational of reality.

The obvious response is, show me. If you can't, I don't believe you.

Brian Romanchuk said...

Tom,

In response to your earlier comment, yes math is just an internally consistent set of statements following from some axioms.

It makes little sense for economists to start solely from axioms,like von Mises wanted to do. As long as you axioms are internally consistent, you can derive a bunch of rules of behaviour for your "economy". But someone else can change those axioms, and generate new rules.

Both model economies are mathematically "correct", but we have to compare to data to see which offers a better fit.

Tom Hickey said...

Agreed.

Conventional economics suffers not only from an excessive penchant for formalism but also from apriorism. Too heavily weighted toward rationalism and not enough toward empiricism.

Matt Franko said...

The "debt to gdp ratio" is "math"...

Ryan Harris said...

The Debate in physics isn't really a parallel to economics. Economics is unknowable because you are looking at a system of rules and laws that aren't fixed, humans make the rules and chose to participate or not according to rules.

In physics the theories are coming up against what is knowable for humans. For example if particles don't interact or exist in the same dimensions where we do, they simply can not be known because by definition our bodies and everything we can interact with only exist in the space and time we see around us. So they come up with obviously false descriptions, like "Big Bangs" or "Inflation," or "Dark Energy" to describe how the dimensions that we exist in came to be but it isn't much different than claiming a biblical genesis because it is all imaginary and known to be wrong at some level, it is simply the best they can come up with to explain what we see given best guesses in the same way the genesis stories came about before there were better tools. And in a few years, new tools will illuminate the unknowable.

Economics is similar in that some things can't be known but the reasons for not knowing aren't physical limitations but rather limitations on what can be gleaned from what is akin to cloud gazing. For whatever reason, economists can't be happy to identify thunder clouds, or storm systems they have to identify, cow shapes, little bunnies, witches and other silly things that mostly exist in the imagination.

But you know, people have to work, bury money in mines so it can be dug up later. So go on, keep measuring pce seasonally adjusted and modified and then transformed with chained inflation, use incomes reported by the tax authorities and then extrapolate productivity to figure out laws of wealth until your heart is content and your pockets full. Good luck with that.