Hacker News Reader

Comment by _wldu | Hacker News Reader

Yep. Economists and meteorologists. What they try to predict is sort of like trying to predict where the disc will land in Plinko [1]. Neither of them have a clue until it actually happens... "Our models showed the disc going the other way (especially the European Model), but we were wrong again. Tune in tomorrow while we talk fast and act confident again."

[1] - https://en.wikipedia.org/wiki/List_of_The_Price_Is_Right_pri...

View on HN

Meteorologists have it easy, because weather doesn't react to the forecast and modify its behaviour to invalidate it.

Self-aware weather with neurosis...

Personally, I always love the "50% chance of rain" forecast as it perfectly describes that they have no idea if its going to rain or not, but yet sounds so scientific and official.

Predicting 50% can be validated: if I say a coin will be heads 50% of the time, I know more than nothing: I know more than someone who incorrectly claims it will be heads 75% of the time.

And we can quantify in various ways how much better the 50% prediction is than the 75% prediction.

That's actually not what "50% chance of rain" means.

Here's a pretty good explanation:

https://www.wsfa.com/2019/08/23/what-does-chance-rain-really...

And many more:

https://www.google.com/search?q=50%25+chance+of+rain

That is absolutely not "we have no idea". It's a very precise prediction, not like a [10%-90%] range. If you look at all their 50% prediction, the half of it should be rain and half should be no rain. This is what this number tells you.

A butterfly flapping its wings would disagree with that.

Meteorologists are pretty good at their job.

within a limited time horizon that is. IIRC meteorological forecasts are about 80% accurate a few days in, 50% accurate about two weeks into the future, and after that of not much use.

Economics of course is often expected to make much stronger claims, people have expectations of it that resemble Asimov's psychohistory despite the fact that there is no science that manages to tame that level of complexity.

I think it was Tyler Cowen who said once that economics is more useful as a tool to clarify thought rather than a tool to make predictions. People just have the wrong expectation of what economics is.

Meteorology is much easier because the act of working out what the weather will do next, won't have an effect on weather. Not so in economics!

My intuition says there's probably a kind of uncertainty principle or incompleteness theorem at play in economics.

It works the other way too. If everybody believes that your model is correct, the chances are very good that its predictions actually become true.

That's the driving force behind a lot of religions and other traditions.

Economics at its base is group psychology, and humans are irrational and wildly unpredictable. A lot of economics writing depends on a rational human that actually doesn’t exist.

Hence why a lot of economics is speculative and should only really be applicable to specific cultures that are being studied, since cultural behaviors and expectations will effect economic behaviors and outcomes.

There’s a balance between data and interpretation in economics, which is why it’s a social science. There are disciplines like Austrian economics that discount quantitative analysis, but that’s as wrong as relying purely on quantitative analysis.

A human is irrational and wildly unpredictable, but groups of people (like most animals) are more easily predicted. While it's difficult to predict a specific stock, it's easier to predict trends given global situations, within a certain time frame, with enough information. The main problem becomes having access to such information. So much of the economy is global now, and behind closed doors, that information is unavailable or at a premium.

> IIRC meteorological forecasts are about 80% accurate a few days in, 50% accurate about two weeks into the future, and after that of not much use.

I think the accuracy is more complicated than that, because while they lose precision (and thereby the correct prediction of the amount of rain in your neighborhood or town), they can predict larger movements pretty accurately pretty far in the future. I remember how bad weather prediction was 30-40 years ago, and am shocked how they now have pretty good insight into the weather two or three weeks from now.

we had have chaos theory. it indicates that we just can not predict things in too far future

>IIRC meteorological forecasts are about 80% accurate a few days in, 50% accurate about two weeks into the future, and after that of not much use.

80% accurate about what ? If this is overall then I'm not impressed because in stable weather you'll get a high % of being right just by extrapolating. Similar with economics for that matter. What I usually care about and where weather forecast failed me just last month, is unstable time over a specific area. Just last weekend my wife and I were about to cancel our trip to my hometown because the forecast was high probability of rain for the entire weekend, this was the forecast on Friday. Went anyway and got a sunny Saturday and Sunday morning, got a bit cloudy by the end of Sunday when we were leaving. This happens so often that I don't know why I bother checking anymore.

Well, there has been a little-noticed revolution in meteorology in the US in the past 15 years, driven by improved simulations and measurement systems (mostly funded by the federal government). Our predictive capability is well over 10x improved from just the late 90s when it was still quite poor.

Not in Seattle. Today was forecast as a dry day (even this morning), but it rained. I tend to do better just by stepping outside, sniffing the air, and looking at the clouds. It's not expertise on my part, just experience.

The problem is economists who say things in the public sphere are heavily scrutinized and criticised and feel the need to 'prove' their theories. Also, because of the pay factor, many great economists go make millions working for banks and hedge funds. Economists know, economics can predict it, but for reasons, it doesn't end up as public knowledge.

Economists are not trying to predict anything. Some economists do make predictions, but economics is not about making predictions.

What is it about then?

“The curious task of economics is to demonstrate to men how little they understand about what they imagine they can design” -- Hayek

Excellent quote

Same as any scientific endeavour, studying a particular phenomenon, which in the case of economics is the production, distribution and consumption of goods and services.

It is in the ability to predict novel situations that understanding is tested.

Not really. For example, a model can accurately explain variable X as a function of Z, Y but if Z and Y are exogenous variables, i.e. outside of the scope of the model, then the future behaviour of X cannot be known, despite the fact that we may understand very well the mechanisms by which Z and Y cause X.

Which is why economics is not a science.

Science is the application of the scientific method, and mainstream economics fully adheres to the scientific method.

There are heterodox schools, such as the Austrian school, that reject the scientific method, and these are definitely not scientific.

Economics is a social science and as such, is not really amenable to the methods of falsification or experimentation that work well in physics or chemistry. In short, it's not really a science. That doesn't mean you shouldn't try to do your best to be logical and factual as much as possible, but at the end of the day you will produce mostly unfalsifiable theorizing.

Really you need more modest goals, for example to at least try to approach the subject objectively rather than via sentimental moralizing. Even that is a massive effort. Imagine a physicist decrying how "wrong" it is for gravity to be weaker than the other forces. You would laugh at such a person and immediately classify them as not a real physicist. So you can use that as a filter to exclude much of heterodox economics and economists as a start. That's the battle being waged -- objectivity -- not falsifiability.

If social sciences are not sciences, what are they? Please, explain. Also I'd like to know how falsifiability applies to the big bang theory. Thank you!

Falsifiability applies to the big bang theory because the big bang theory is not about the general concept of "a" big bang, but about specific models of how exactly the events very shortly after the big bang happened, which make testable predictions about all the consequences of those very early events - the distribution of matter and energy in the universe, properties of cosmic microwave background, etc; such models can predict observations/measurements (of ancients events) that we have not yet made, and be falsified if those observations disagree with the model predictions.

Falsifiability is feasible to some degree -- e.g. in history you can hypothesize that a civilization had copper and then look for evidence of copper artifacts.

But in the physical sciences you have well-formed mathematical models that make very specific and precise predictions and can be tested to a high degree of accuracy. That specificity allows you to set threshholds like 5 sigma for acceptance of a result as being verified.

"Did we find copper" is a bit different, and thus there is a fuzziness in history that you don't have in physics or chemistry, which is why we don't view history as being a science even though it benefits from use of the scientific method. Everything -- cooking, dating, gardening -- can make use of the scientific method, but at issue here is the trust that the public places in accepted findings of science and such trust is not justified in the fuzzier fields. People need to add big error bars in their heads when listening to results from social sciences and much smaller error bars when listening to results from physics.

A good rule of thumb is that fields which style themselves by adding "science" to their name are not sciences. An economist friend of mine told me that, and it stuck. In fact if you are a physicist you don't go around insisting to people that you are a scientist and what you do is science and if they disagree with you they are anti-science. You just say "hey, this is physics".

So far, this rule of thumb has not let me down.

FYI, you may want to read Feynman's lecture on Cargo Cult Science -- that is avoiding the trappings of science while not actually having the rigour and intellectual honesty of true science. A good link is here: https://calteches.library.caltech.edu/51/2/CargoCult.htm

I enjoy economics, political science, anthropology, sociology, etc, but they are most definately not science.

> good rule of thumb is that fields which style themselves by adding "science" to their name are not sciences

And this is your whole argument?

Isn't physics a "natural science" and therefore not science according your rule of thumb.

This discussion is stupid. If you want to argue that social sciences are not science, all you have to do is show that social sciences don't follow the scientific methodology. If you can't show that, you should shut up and stop wasting everybody's time.

You can't really prove or disprove whether an economic theory is a good choice or not for a country at a specific time.

How are you going to measure the outcomes and for how long will you measure them?

The problem with economics is how politicised this field is and how the best theory ends up being what's best for governments and politicians. Is mainstream economics what's more convenient for the 1% or what's best for the world?

We will never know what the economy would be if we weren't so Keysian. Maybe we wouldn't have a crisis every 10 years.

Your model of X, Y, and Z cannot make predictions. How do you validate it?

Generally speaking, statistical models are validated by measuring how well/bad they fit the data, for example, studying the residuals and performing statistical tests. Predictive power can be estimated by using the model to predict known outcomes (using out-of-sample data).

"To most people, economics is a dull science full of statistics and jargon, mainly concerned with money and designed to answer a narrow (but important) set of questions. To economists, economics is a powerful tool for understanding why armies run away, voters are ignorant, and divorce rates rise, as well as solving practical problems such as how not to get mugged. Its theme is not money but reason- the implications, especially the nonobvious implications, of the fact that humans act rationally. Or to put it more formally:

Economics is that way of understanding behavior that starts from the assumption that individuals have objectives and tend to choose the correct way to achieve them."

Excerpt from Hidden Order, a book which explains economic concepts to non-economists.

Explanations are useless if they can't make predictions.

Non-falsifiable explanation aren't worth the paper they're printed on.

While it's a nice quote above, I suspect many economists would disagree with that reduction of their field to being only backwards looking.

> Explanations are useless if they can't make predictions.

Nonsense. Most sciences aren't about making predictions. Another example is linguistics. Linguists have reconstructed languages long extinct and figured out how they evolved into modern languages. Yet, they can't predict how these languages will evolve in the future. Does this mean these explanations useless? No. Why should they be useless? They may not be of interest to you, but that doesn't mean they're useless.

I disagree with you, strongly, but I'll just note this -- you appear to espouse the view (in you last sentence) that utility is the only value something can have. There are many things to know that lack utility but still hold value somehow. Useless does not mean worthless; but science really is about prediction.

I don't know what you're on about. It's been over a century, since economists adopted a subjective theory of value, according to which value is that—subjective. As such, we don't try to explain it, because it doesn't matter. You may think something is valuable because it's useful, or because some other reason. It makes no difference, and we don't care. All we care about is how much you are willing to pay for it. This tells us everything we need to know.

That's pretty arguable. There are many things that I find useful that I wouldn't pay for.

It's mainly about allocating resources.

Of course it's about prediction. There's no point spending time trying to understand something unless you're going to use that understanding to try and anticipate the future.

What I find amazing is that people can make careers (Economists and Meteorologists) out of guessing and being wrong most of the time. Anyone with a real job would get fired within the first month if they had the same record of performance.

Excellent.

Reply on news.ycombinator.com