Econometrics: What is Econometrics?

hello name is Nick Huntington Kline this is the first in a series of videos on econometrics and so the first one is just going to be what is econometrics what are we doing here that's the whole point so econometrics is the way that economists work with data and we try to learn things from data now there's a lot of people who try to learn things from data statisticians there's data scientists but the goals that these different groups have are slightly different so the way that we work with data is a little bit different so take a data scientist for example a data scientist is largely interested in making good predictions right so they want to say ok well we have all this information about a customer what's the probability that they're going to purchase something right that might be a good data science task or I have this picture I can see all the pixels in the picture based on the pixels in the picture can I predict whether it's a picture of a cat or a picture of a doll another good data science task what economists are mostly interested in is not so much prediction right people tend to think that economists are all about like predicting what the future of the economy is gonna be and there are a few economists who do that but really most of the time that's not what we're doing what we are interested in doing is getting very accurate estimates of the relationship between different variables so we're not so much interested in let's say trying to predict what GDP is gonna be next month although again there are people who try to do that we're more interested in let's say how does our tax structure affect our GDP growth right we want to know the relationship between tax structure and GDP growth and we want to get a very accurate idea of what that relationship is and not just the relationship that we've seen in the past but what is the underlying true relationship between those two things right in a counterfactual world where let's say we could change the taxes and to be something a little bit different how would you grow be different right that's something that's very difficult to answer when you're just looking at raw data because the raw data that you have all came from the world that actually exists and so if you're trying to get something from the world that does not quite exist yet things get a little bit thorny err so there's actually a pretty easy solution to this problem where we're trying to get a very clean relate an estimate of the relationship between two variables and that is running a randomized experiment if you could randomly assign the tax structure you could very easily figure out how tax structure affects GDP because you just see okay when I assign the tax er should be like this this is what happened to GDP when I design the tax structure to be like that that's what happened to GDP but we can't do that a lot of the questions that economists are interested in you can't run an experiment either because it would be impossible you can't actually get governments to agree to just randomly assign their tax start just like an economist wants them to or because it would be infeasible maybe you could run an experiment with super-expensive or it could just be immoral right you know you can't randomly assign people to do something that's harmful for them just so you can see what the effect of that thing might be so we can't run an experiment and again there are economists who run experiments that's not what this is about econometrics is about taking data that already exists observational data that was collected not for the purgin without the control of the researcher right the data was collected outside the control of the research maybe they were in charge of the data collection process but they had no say in what actually went on there they couldn't randomly assign a treatment all they could do is look and observe and see what's going on in the world then taking that observational data and drawing inferences making making claims about the underlying data generating process that went into that data so what is the true relationship between two variables that we can figure out based only on looking at observational data that's what econometrics is all about now for doing this economists have to think a lot about what's called the data generating process now the data generating process is whatever set of underlying true laws there are that make us see the data that we see so for one example let's say that I have something in my hand right and it's up here and then I drop it and it's down there right what are the observations that I have well I observed then I was holding it and it was up here and then I also observed that I dropped it and it was down there right so those are my observations what's the data generating process that led me to see those observations well the law of gravity would be part of the data generating process there right it is why when I dropped the nail clippers they dropped down right that's that's the sort of why of what's going on right think of a data generating process as why why do we see the data that we see why do we see the relationships between variables that we see and we need to think about the data generating process very very carefully because without the ability to run an experiment we need to take advantage of what we think we know about the data generating process to learn something new so let's take a quick example and we can just see the kinds of tasks that we're up against and how difficult they can be and then of course the rest of this class will be about how we can solve the obvious problems that are going to pop up here so let's talk about an example of data generating process and one that should be familiar to you if you've taken economics before which is the model of competitive supply and demand so as economists we would say ok if we have a competitive market let's just say we have a competitive market we observe the prices and quantities might change from time to time sometimes the price is higher sometimes it's lower or sometimes more units get sold some times fewer why why do we see the prices and quantities changing around in a competitive market well an economist will tell you it's probably because of supply and demand shifting around so that's our theory that's our data generating process we would say that the reason we see the prices and quantities that we observe is because of the supply and demand forces working underneath the table right now it's our data generating process and we can map this out so here's a standard graph of supply and demand and you can see our theoretical supply line going sloping up there are theoretical demands looking down and what we get from this is a single observation we the only thing that we actually see is that red dot right there we see that the price is 15 we see that 50 units were sold that's what we see the theoretical supply and demand curves are invisible to us we are theorizing that they are there and our goal as econometricians is to try to take those observations take just the red dots and figure out what the black lines look like so we're trying to use our observations despite the fact that we can't run an experiment to learn about the data generating process use the prices and quantities to learn what supply and demand look like and that's just one example right it's not all supplying them in the prices that we see they're coming from this theoretical construct of supply and demand given that theoretical construct and the observation that we can see if we do our economy tricks we might be able to work to learn something about the data generating process that were actually interested in and this will help us learn more economics because economics has to do with these theoretical structures that we think there and we want to learn are those the right theoretical structures are they the wrong ones can we show that they're the wrong ones if for the right ones one of the specific details we if we think the supply and demand are going on how steep is supply or demand right that's something that we could figure out from our observations of course this is difficult right it's not always obvious how to take an observation and turn it back into a theoretical claim and this is a problem that we're gonna come up against again and again so let's say for example these are the observations that we saw we followed the same market for five weeks in a row and these are the prices and quantities that we saw how could we take these observations and turn them back into supply and demand curves right I could say something like oh yeah when we move from this point right here to this point right here that must have been because Oh demand shifted left how do I know it's not because supply shifted right and so each of those claims would lead me to a different conclusion about the slope of supply or the slope of demand if I was trying to learn about those two things and so being able to figure out which of those two is going on and how we can isolate using statistics just one of those explanations which is the process called identification that we'll talk about as well that is the sort of main task of econometrics so what are contracted what's econometrics all about it's all about trying to precisely estimate the relationship between two variables and not just precisely estimate the relationship that we see but precisely estimating the theoretical relationship that's running under the hood in the data generating process when we do this it means that we are trying to take observations that we actually see and use them to learn about theoretical structures which can be difficult because as you can see on this graph right here it's not always obvious how you can take a set of observations and turn those into something that you can know about a theoretical structure and this whole class is going to be figuring out how we can do that when we can do that and women can't thank you

Channel: Nick Huntington-Klein

Views: 1,144

Rating: 5 out of 5

Keywords:

Id: 1-YDYB917_k

Length: 8min 55sec (535 seconds)

Published: Wed Jun 17 2020