Is Studying Economic Data "Doing Economics"? \ stacker news ~econ

Is Studying Economic Data "Doing Economics"?

3192 sats \ 43 comments \ @Undisciplined 6 Feb econ

Studying Economic Data Is Not "Doing Economics" By Frank Shostak

Economists are fond of claiming that employing data and statistical analysis is actually “doing economics.” No, they are “doing data” and nothing more. Real economics employs real theories that explain economic phenomena.

I largely agree with the sentiments expressed in this article, although I would frame some of it differently. I also don't really like the gatekeeping mentality expressed in the title.

There are plenty of economic questions that are inherently empirical, so studying the relevant data using appropriate statistical methods is "doing economics". That isn't at odds with anything in the article, provided the hypotheses being tested are rooted in philosophically sound theory.

view all related items

748 sats \ 5 replies \ @SimpleStacker 6 Feb

Yes, it is "doing economics". To say otherwise would be to render the entire field of economics unfalsifiable. Even historical analysis is a kind of data analysis, except you're doing a deep dive on a handful of data points rather than looking for statistical regularities across a large number of data points.

It's true that economists aren't measuring unchanging laws of nature like a physicist would. But the underlying assumption is that there is enough regularity in human behavior that these quantitative relationships are useful for making predictions about the future or for testing the validity of economic theories.

105 sats \ 4 replies \ @Undisciplined OP 6 Feb

We've touched on this elsewhere, but I don't think it's right to say that theory needs to be empirically testable to be falsifiable.

In the same way that mathematics is falsifiable, so are deduced economic theories. Finding a contradiction is how that kind of scientific theory is falsified. As with mathematics, there is also the question of whether the assumptions that theories are based on are appropriate for the situation.

656 sats \ 3 replies \ @SimpleStacker 7 Feb

Fair enough. I retract my statement about falsifiability. Still, even a solid economic theory often requires empirical measurements in order to make predictions, e.g. supply and demand elasticities. In that sense, I would still argue that data work is an indispensable part of economic analysis. I get where the author is coming from, but I also think it's a bit pedantic. Does anyone really ever claim to do economic data analysis apart from a theoretical framework, even if the framework is implicit?

25 sats \ 2 replies \ @Undisciplined OP 7 Feb

There are people who don't realize they're bringing theory into their data analysis. You might here them say things like "let the data speak for themselves". They fancy themselves as being ideologically neutral, but they tend to just not understand the ideology they're implicitly assuming.

I get where the author is coming from, but I also think it's a bit pedantic.

I agree

a solid economic theory often requires empirical measurements

also agree

444 sats \ 0 replies \ @SimpleStacker 7 Feb

True. And I often see economists come up with a crazy model that, with the right choice of parameters, can fit the data and ergo "the model is supported by the data". Well, no, if the model is inherently flawed then it doesn't really matter if it can fit the data, it's not going to be useful for predicting or understanding the world

0 sats \ 0 replies \ @Undisciplined OP 7 Feb

*hear

I hate catching typos after the edit window is closed.

728 sats \ 1 reply \ @grayruby 6 Feb

Doing economics is such an odd term but wouldn't aggregating economic data for the purpose of analysis be "doing" economics similar to the way collecting the resulting data from an experiment would be "doing" science?

100 sats \ 0 replies \ @Undisciplined OP 6 Feb

I don't like the term either. The point of the article is that unless the work is guided by economic theory it is not "doing economics". Data work in service of demonstrating an economic relationship would be part of "doing economics", I think.

I'd say the author was trying to emphasize that hypothesis formation in economics is different than in the natural sciences. In economics, theory comes first, whereas in the natural sciences observation comes first.

696 sats \ 5 replies \ @cryotosensei 6 Feb

On first impulse, I felt that Frank Shostak is splitting hairs. Isn’t the collation and curation of data part and parcel of the process of testing a hypothesis? I thought he was thinking of doing data as grunt work and something beneath his level.

Upon further thought, I am thinking maybe he wants to distance his work from the bare bones of data collection. With the prevalence of LLM and AI tools out there, perhaps the human input involved in data collection is weighed down by comparison, so he doesn’t want his work to be undermined as a result

129 sats \ 4 replies \ @Undisciplined OP 7 Feb

Isn’t the collation and curation of data part and parcel of the process of testing a hypothesis?

Not in the Austrian tradition. This is a fundamental point they make about how economics differs from the natural sciences. Instead, economic theory is inherently axiomatic and follows deductively from those premises. In that way, economics is more like mathematics (or is a branch of mathematics) than what people generally refer to as "science". (Although, both econ and math are sciences.)

444 sats \ 1 reply \ @SimpleStacker 7 Feb

I appreciate the Austrian perspective greatly, but I confess that I am not well schooled in their thinking. How do Austrians handle situations in which a quantitative measurement is needed to make a prediction, like a demand or a supply elasticity? Do they not consider that part of economics proper, like how engineering is considered separate from physics?

10 sats \ 0 replies \ @Undisciplined OP 7 Feb

like a demand or a supply elasticity

This is actually my main gripe with the way they talk. Estimating elasticities is absolutely part of economics proper and they acknowledge that. It's easy to miss that, though, when reading pieces like this, even though the author never actually says anything to the contrary.

like how engineering is considered separate from physics?

There are a couple of interesting parallels to that.

Finance, even if rooted in Austrian theory, is usually considered a separate discipline from economics.
While still being considered part of economics, Austrians have an interesting perspective that empirical work is best understood as doing economic history.

133 sats \ 0 replies \ @cryotosensei 7 Feb

Today I learnt! Haha

133 sats \ 0 replies \ @TNStacker 7 Feb

The Austrians preceded the behavioral economists!

674 sats \ 4 replies \ @shado_op 6 Feb

"All that quantitative methods can do is describe movements of historical data; it cannot identify the driving forces of economic activity." Being honest, I'm not sure I get the point... If an area loses it's main employer, with no equivalent replacement, quantifiable outcomes can be observed and measured. Unemployment, under-employment, reduced tax revenue, ill health, crime, family break up, drug and alcohol addiction often follow. (Eg, Rust Belt in the USA, North East in the UK). The local driving force of economic activity is the loss of jobs and the impact can easily be measured. So what am I missing here?

25 sats \ 3 replies \ @Undisciplined OP 7 Feb

Good question, there are two things I notice:

As obvious as it may be, you are proposing a theory of the connection between a specific employer and downstream effects, so it isn't just the data identifying the driving forces. That's often referred to as the "theory laden" problem in science. Without theory of some kind you'd just be grasping at random straws.
A major part of economic analysis involves understanding both what was observed and what wasn't (the counterfactual). This is the "compared to what?" piece. In your example, the employer is going out of business for a reason, so you'd need to know what the alternative was to them going out of business in order to make draw meaningful conclusions.

444 sats \ 2 replies \ @shado_op 7 Feb

Thank you. So the Mises approach is you need a theory to help you understand the data? I get that from a philosophical pov, but the problem is the natural human tendency to goal seek. Many of us see this all the time at work. The 'guys at the coal face' can tell what is going wrong based on observation and heuristics. But someone in authority will ignore this practical experience, based on the data; in other words, the data has been mis-interpreted in line with a theory, which blinds the manager/analyst/consultant to reality on the ground. Seems to me we should analyse the data first, then reach a conclusion. My example of a big employer closing is deliberate to illustrate the point. When UK/US manufacturing started to collapse in the 70's and 80's, many pundits claimed this was a positive economic development, and there would not be long lasting issues. They were very wrong, imv, though I understand the opposite arguments. But the data evidencing the fact such areas were plunging head first into catastrophe was ignored, because of economic theories. Btw, I agree with a lot of the Austrian School articles! But on this point, I feel data should be analysed free of bias, and that includes political & economic theory. It should be data first, theory second to explain the data...? Good debate on this post btw!

100 sats \ 1 reply \ @Undisciplined OP 7 Feb

There's a lot to unpack there. The feedback cycle between theory and practice is very interesting.

I would expect the Austrians to give more weight to the on-the-ground experience of people who do the work than the empirical schools of economics. That's an ironic reality, but the empiricists will dismiss what they're told with "the plural of anecdote is not data". In other words, workers' observations are not gathered in an experimentally rigorous way, which leads empiricists to simply dismiss them.

An Austrian would more likely appreciate that work processes evolve over time to reflect economic reality and changes observed by workers first-hand are evidence that economic reality may have changed.

One of the core tenets of the Austrian School is epistemological humility, which means being humble about what you claim to know. They recognize that economics is too complex to be modelled precisely, so they aren't going to make the kinds of overreaching claims that neoclassical or Keynesian economists make.

I feel data should be analysed free of bias

The technical meaning of "bias" is "a tendency to make errors in a particular direction". It isn't about coming into an investigation with prior beliefs. If your beliefs are true, then you won't be biased. If you come in with no prior beliefs, you won't have any idea what questions to ask.

To me the Austrians are by far the least biased school of economics, because they make the fewest simplifying assumptions in their theory. All those economists you were talking about being dismissive of how local economies would be affected by manufacturing loss were overconfident in the assumptions in their models. Those assumptions are the source of their biases, but they would tell you that those assumptions were backed by data.

Good debate on this post btw!

I agree! It's a lot of fun for me.

305 sats \ 0 replies \ @shado_op 7 Feb

Good points, Cheers!

650 sats \ 3 replies \ @TNStacker 7 Feb

It is a part of it, but not the whole discipline, which means the original question may be faulty, at worst, leading at best. Did the writer mean a part of doing economics? The way the question is written leaves if fair to conclude they meant all. To that the answer would be no.

21 sats \ 2 replies \ @Undisciplined OP 7 Feb

To be fair, the author's title was Studying Economic Data Is Not "Doing Economics". I made it a question for the post, because I thought that was better for sparking conversation.

133 sats \ 1 reply \ @TNStacker 7 Feb

It was a very fine question then, sir.

0 sats \ 0 replies \ @Undisciplined OP 7 Feb

Thank you

653 sats \ 15 replies \ @nkmg1c_ventures 6 Feb

I listened to the audiobooks of Mises' Human Action and got through most of it (probably should go back), he talks about this.

After 1900 with WWI and particularly WWII to follow, the use of math and science in warfare became very important. So there came about a class of professional scientists and engineers who dealt with everyday problems of logistics, munitions, medicine/health etc. A lot of this was very data driven, but also theory driven. Graph theory for example might have seemed a mental curiosity of mathematics before maturing with its applications in warfare, logistics, computing.

What Mises and other Austrian-influenced economists are railing against is the use of bunk statistics to justify theories of money and human behaviour that don't make sense. They are distrustful of coming up with Keynesian-style models of inflation that justify certain human behaviour (Cantillon effect, kleptocracy etc) where the academic supporters are essentially "fudging the data" to support Keynesian theories of monetary policy and taxation etc.

This larger divide played out even in the field of math and statistics more specifically, with many real world applications and results coming of black-box statistical fiddling. The kind of theory-driven thinking of Austrian economics would find this suspect and iirc, Mises actually had a brother (Richard Von Mises) who was a rocket scientist and they often disagreed on these philosophical questions about how to explain natural world, human behaviour. In particular, there was a camp of academics in physics and chemistry but also sociology who thought all we needed is sufficient data fed into some model arrived at statistically, and then we have explained any phenomena we can dream of.

Mises (Ludwig, not his brother) and subsequent Austrian econ guys likely found this line of thinking incredibly suspect because they'd already seen it abused with monetary policy.

In order to continue the modern system of constantly debasing money, blaming it on high employment, firing people cyclically and printing more money -- in order to continue this foolishness, you basically have to hide yourself in data and spreadsheets and ignore the underlying system / mechanics. I think that's what's at play with this mental divide and way of thinking about economics.

Someone who's done a lot of stochastic math and stats should feel comfortable fitting models and discussing data, IMO, but shouldn't be married to anyone model structure because that is often where the "faith" and "fudging the numbers" can be hidden, in spurious variables and parameters.

My $0.02

Richard Von Mises -- https://en.wikipedia.org/wiki/Richard_von_Mises

66 sats \ 14 replies \ @Undisciplined OP 6 Feb

What Mises and other Austrian-influenced economists are railing against is the use of bunk statistics to justify theories of money and human behaviour that don't make sense.

Great comment! I just wanted to add to this part that they are also railing against the type of deterministic mathematical models used in physics. Those models inherently rely on fixed parameters (constants of nature) that have no analogue in economics.

I think they are also correct about that point, but I also think it is the source of an overzealous rejection of mathematics in general.

1534 sats \ 9 replies \ @nkmg1c_ventures 6 Feb

Mises, iirc, didn't say such constants are non-existent. He has an interesting aside where he talks about the limitations of human knowledge and the effect this has on human knowledge. He has asides about "super-intelligences" and kind of "god mode" views of a phenomena, where he admits perhaps future man would be able to create theories with more perfect information (think mass computing, data surveillance, etc, my emphasis not Mises').

But he was very adamant that the approach of his contemporaries was confused because of the way they tried to jam physics approaches onto questionable models of finance, because he saw it wasn't true and didn't reliably work. There is also the issue of defining units when transferring the physics approach to finance, which Mises rightly saw as a problem that physicists would be appalled at. Imagine not keeping track of meaningful units as a physicist lol, unheard of.

21 sats \ 8 replies \ @Undisciplined OP 7 Feb

the issue of defining units when transferring the physics approach to finance

My undergraduate background was physics and mathematics. When I went to grad school for econ, the reckless disregard for units in econ models blew me away. I had already been skeptical of that kind of modelling from reading the Austrians, but it would be jarring to anyone approaching it from physics or chemistry.

274 sats \ 7 replies \ @SimpleStacker 7 Feb

Curious what you mean as to the reckless disregard for units. I have a whole bunch of problems with how economists work, but disregarding units wasn't on my list of complaints.

640 sats \ 4 replies \ @nkmg1c_ventures 7 Feb

The units thing becomes apparent when you try to bridge modern micro and macro economics. Responsible econometrics people use units as best they can, but in order to bridge theories to make them intelligible you get into a lot of synthetic units that only make sense in certain models. There is also a classic criticism from Austrian school that the way that decisions are made at the margin are often ordinal -- meaning we rank goods and costs according to some rank/order, but we can't meaningfully assign cardinal units (number units) to them.

The other side of this argument is we now have computing power and mathematical theory we didn't back then in Mises' time to help us mixing ordinal / cardinal models, but the other side (modern econ) really likes to stick to black-box models of inflation that obscure actual realities. Think CPI core inflation (minus housing and energy)... a joke.

444 sats \ 2 replies \ @nkmg1c_ventures 7 Feb

Another way to put this would be that modern quants are using a variety of models that work and provide predictive and hopefully sometimes explanatory power, but that comes from sophisticated models that bear no allegiance to economic dogma. They're just models that survive and are useful.

20 sats \ 1 reply \ @Undisciplined OP 7 Feb

That reminds me of reading about the development of artillery targeting formulas. As one might expect, they started with the simple parabolas suggested by classical mechanics, but those turned out to be incredibly inaccurate, because none of the simplifying assumptions hold. Then, they tried to add in the frictions and other factors, but that became too complicated. Ultimately, they solved it through massive trial and error, where they explored how dozens (maybe hundreds) of variables affected the shell's trajectory.

Even when all the relevant theory is precisely developed, there can still be value in doing atheoretical empirical work.

To tie it back to the original prompt, though, were all those artillery specialists "doing physics"? I think it's reasonable to say they weren't.

view replies

0 sats \ 0 replies \ @nkmg1c_ventures 7 Feb

deleted by author

31 sats \ 1 reply \ @Undisciplined OP 7 Feb

There are lots of mathematical models that are used by economists where the units just don't make sense: for example, you might see an exponent that has units. I remember being surprised by this, because in physics checking to make sure your units make sense is routine.

0 sats \ 0 replies \ @nkmg1c_ventures 7 Feb

deleted by author

665 sats \ 1 reply \ @ek 6 Feb

I want to contribute to this discussion by mentioning this great quote:

All models are wrong, but some are useful.

— George E. P. Box

1 sat \ 0 replies \ @Undisciplined OP 7 Feb

That's one of my favorite quotes and I had no idea who said it. Now, I still have no idea who said it, but I do know his name.

665 sats \ 1 reply \ @nkmg1c_ventures 6 Feb

I think it's a bit more subtle than that.

If you consider that exponential functions are inescapable in financial mathematics and modeling human economies, it could be argued certain constants like 'e' and 'pi' count as esoteric constants of economics -- however I understand what you mean. They are dragged in almost through the universality of mathematics itself, and we can express many functions using these constants without them having any explanatory nature.

I'm still not certain such constants don't exist for econ -- constants which describe limiting behaviour in graph theory (human relationships, barter, etc) exist and so may naturally be considered "constants of finance/economics".

But yes, the whole modern Keynesian / mathematical econ black box where they pretend the theories about inflation and the variables in them are akin to proven physical Theories (theories with the capital 'T' in science), it's not intellectually sound.

30 sats \ 0 replies \ @Undisciplined OP 7 Feb

Totally fair correction. I was thinking about what you discuss towards the end.

305 sats \ 0 replies \ @kurszusz 7 Feb

Studying economic data make you a statistician, not economist. To be an economist it is much more that to study and analyse data / statistics. Yes, I agree the statistics are part of economy...but economy covers a much larger area

269 sats \ 0 replies \ @Entrep 7 Feb

Studying economic data can do by any data analyst, doing real economic involves applying data to learn to policy making such that it affect economic growth in decision making

248 sats \ 0 replies \ @phatom 7 Feb

This studied data can help economics so I will say it is doing economics

0 sats \ 0 replies \ @nkmg1c_ventures 7 Feb

deleted by author