I think sciences should really be classified on degree of complexity of the phenomena they're dealing with.
Physics is a simple science. It studies the most basic of phenomena, and can isolate them pretty well for experiments. Chemistry is a more difficult science, biology even more so - at each step the phenomena get an order of magnitude more complex (in computational sense), and thus more difficult to study. Then there's psychology, social sciences and economics - all of which are even more complex, as they no longer deal with lots of simple machines, but results of sentient minds interacting in the complex world.
Point being, the "softer" a science is, the more difficult it is. That the "hard" sciences spawn so clear predictions and mathematical models is a consequence of their simplicity. The proper approach here is humility - economics and social sciences are so difficult that we barely begun to figure anything out. That's also why there's so much bullshit in "soft" sciences - they're so difficult that you can plausibly say almost anything, and it's hard to tell if you're wrong, or how much.
I don't think that's true at all. Pharmacology is a very difficult science that we get wrong all the time, yet we still have medicine because we are able to test our medicines by trying them on animals and people. Likewise in economics and sociology and astrophysics we are able to test through natural experiments. A test of a model is its ability to predict. We don't accept untested models just because they're "complicated".
And sure, maybe you believe that economics has more untested models, but that doesn't mean there aren't many tested models in economics.
Edit: reply button is locked out for some reason so replying here.
>And look at how much it costs, how unreliable the results are, and how even after full batch of successful trials, we have no clue why a drug works.
And yet we still do solid pharmacology that results in reliable medicine, just like we can do solid economics that results in models with high predictive power.
>If you take a large enough sample and apply enough statistics to it, you can see something that looks sort of like evidence to your hypothesis. In the end, you're not much closer to knowing why it happens (and why it happens only some percentage of the time). Compare that to hard sciences, where most of the time, you can design experiments based on first principles, and you can connect the results back to fundamental theories through a string of formal math.
The idea of statistical approximation does not distinguish between controlled experiments and observational study at all. We were able to statistically model the relationship between speed and energy before we could explain gravity, i.e. we were able to figure something out before we could explain why. And most experiments in physics resulted in approximation, such as newtonian physics, that works well enough at low speeds but breaks down at high speeds. Our current model of gravity was made in the 1970s, yet before 1970 we could explain that when you jump you fall to the earth. The fact that approximation is used is not remarkable in the slightest. That's what modeling is. A model that is too complicated is useless, a model that is too simple is inaccurate. All models are approximations to some level.
>Again, "tested" in physics vs. "tested" in economics are apples-to-oranges.
Yeah, that's why newton discovered relativity, right? No, we have areas of clear evidence and areas outside of our ability to measure for every field. Just becomes some areas of a field are difficult doesn't mean every area is difficult. Just because we couldn't model what happens to a ball thrown at 0.9c in the 1700s, doesn't mean we couldn't model a ball thrown at 0.9 m/s.
>Ultimately, my main point is that economics is so hard that even our best results are pretty bad, and marginally useful.
Look on the IGM site for policy recommendations that we have clear economic consensus on. ]If a country chooses to ignore economics, e.g. zimbabwe and greece, the results are very predictable.
> Pharmacology is a very difficult science that we get wrong all the time, yet we still have medicine because we are able to test our medicines by trying them on animals and people.
And look at how much it costs, how unreliable the results are, and how even after full batch of successful trials, we have no clue why a drug works. This is because medicine, as a practical ofshoot of biology, is a field so complex, it defies reasoning from first principles. We don't have a full model of how things work, so we rely on the "shotgun approach" - statistics.
Same thing happens in psychology, sociology and economics. If you take a large enough sample and apply enough statistics to it, you can see something that looks sort of like evidence to your hypothesis. In the end, you're not much closer to knowing why it happens (and why it happens only some percentage of the time).
Compare that to hard sciences, where most of the time, you can design experiments based on first principles, and you can connect the results back to fundamental theories through a string of formal math.
> A test of a model is its ability to predict.
That I 100% agree with. And it reinforces my point. Hard sciences gives you models you can rely on. Soft sciences don't. Not because they are full of dumb people, but because the problem domains of soft sciences are orders of magnitude more complex than those of hard sciences, and all your models end up being crude approximations.
> And sure, maybe you believe that economics has more untested models, but that doesn't mean there aren't many tested models in economics.
Again, "tested" in physics vs. "tested" in economics are apples-to-oranges.
Ultimately, my main point is that economics is so hard that even our best results are pretty bad, and marginally useful.
Reply button has reappeared so copying my response here for proper flow, but now I can't remove it from my other comment because the edit button disappeared.
>And look at how much it costs, how unreliable the results are, and how even after full batch of successful trials, we have no clue why a drug works.
And yet we still do solid pharmacology that results in reliable medicine, just like we can do solid economics that results in models with high predictive power.
>If you take a large enough sample and apply enough statistics to it, you can see something that looks sort of like evidence to your hypothesis. In the end, you're not much closer to knowing why it happens (and why it happens only some percentage of the time). Compare that to hard sciences, where most of the time, you can design experiments based on first principles, and you can connect the results back to fundamental theories through a string of formal math.
The idea of statistical approximation does not distinguish between controlled experiments and observational study at all. We were able to statistically model the relationship between speed and energy before we could explain gravity, i.e. we were able to figure something out before we could explain why. And most experiments in physics resulted in approximation, such as newtonian physics, that works well enough at low speeds but breaks down at high speeds. Our current model of gravity was made in the 1970s, yet before 1970 we could explain that when you jump you fall to the earth. The fact that approximation is used is not remarkable in the slightest. That's what modeling is. A model that is too complicated is useless, a model that is too simple is inaccurate. All models are approximations to some level.
>Again, "tested" in physics vs. "tested" in economics are apples-to-oranges.
Yeah, that's why newton discovered relativity, right? No, we have areas of clear evidence and areas outside of our ability to measure for every field. Just becomes some areas of a field are difficult doesn't mean every area is difficult. Just because we couldn't model what happens to a ball thrown at 0.9c in the 1700s, doesn't mean we couldn't model a ball thrown at 0.9 m/s.
>Ultimately, my main point is that economics is so hard that even our best results are pretty bad, and marginally useful.
Look on the IGM site for policy recommendations that we have clear economic consensus on. ]If a country chooses to ignore economics, e.g. zimbabwe and greece, the results are very predictable.
> Ultimately, my main point is that economics is so hard that even our best results are pretty bad, and marginally useful.
That’s a very interesting point. Most would consider Econ a “soft science” because it is less like physical science e.g. Chemistry. Weird how people like to flip the script for some sorta political gain. Yes, in the soft sciences it is more difficult to reach conclusions as compared to fields like biology, but should we say that makes them more harder?
Ooops, I tried so much to avoid using word "hard" for multiple meanings, and I failed.
What I wanted to say is that economics is a "soft science", in the same sense physics is "hard science". Also, that economics is a difficult science dealing with complex phenomena, in the same sense that physics is an easy science dealing with simple phenomena.
Physics is a simple science. It studies the most basic of phenomena, and can isolate them pretty well for experiments. Chemistry is a more difficult science, biology even more so - at each step the phenomena get an order of magnitude more complex (in computational sense), and thus more difficult to study. Then there's psychology, social sciences and economics - all of which are even more complex, as they no longer deal with lots of simple machines, but results of sentient minds interacting in the complex world.
Point being, the "softer" a science is, the more difficult it is. That the "hard" sciences spawn so clear predictions and mathematical models is a consequence of their simplicity. The proper approach here is humility - economics and social sciences are so difficult that we barely begun to figure anything out. That's also why there's so much bullshit in "soft" sciences - they're so difficult that you can plausibly say almost anything, and it's hard to tell if you're wrong, or how much.