Jump to content
Urch Forums

richard

Recommended Posts

Just want to check what you guys think about the application of data mining/machine learning/deep learning and other AI techniques in economics field.

I heard that some researchers have been doing this, but haven't looked at their work yet. Also wonder if any PhD programs pay specific attention to this track?

 

Thanks!

Link to comment
Share on other sites

 

Susan Athey is a good start. OP, what exactly are you interested in with regards to ML in economics? You won't find economists developing new ML algorithms, but you will find a variety of applications.

 

(1) Some such as Chernozhukov at MIT are interested in areas like dimensionality reduction and using ML to estimate nuisance parameters. There is an area which probably begins with Chernozhukov at the top interns of rigor, and which focuses on deriving traditional econometric properties of estimators that are fully, or augmented by, ML-based. Simulation and performance are also considered.

 

(2) There is another area of researchers that are simply interested in what it would be like to replace traditional econometric estimation with ML. This is less likely to get you very far for a lot of reasons (chiefly: causality still rules the day in academic economics). See, for instance, Bajari et al. throwing 8 different ML methods at a traditional demand estimation problem and benchmarking out-of-sample performance against traditional IO models (there is some straw-manning happening here, but it's not that bad).

 

(3) Finally, there is another group of empirical researchers who don't "do ML" per se, but are interested in using ML for creative purposes such as unlocking new data sets, augmenting additional data sets, or benchmarking their econometric models in terms of out-of-sample predictive performance. No one here is involving ML at all in their statistical results of interest, they're just using it as a step in the process before the real econometrics are done.

 

These are, at least, the main categorizations as I see them right now in academia. I see (3) as being the most accessible and the easiest to get a foothold in, (1) as the most challenging and the least likely to be feasible unless you're in a top econometrics department, and (2) to be worth our while as a field to take a stab it--if nothing else out of curiosity, but very crude at this point and unlikely to look good as the core of a dissertation or JMP.

 

Economists working at places like Google and Amazon will be stepping even further outside of these boundaries. I'm personally working on combining aspects of (1) and (3) above. I'd be happy to point you to more resources if the area interests you. I'd say you need a strong foundation in econometrics, as well as coding chops and some computer science background. Beyond that, outside of the top 5, working in ML will require you to be creative with the resources at your disposal and you'll have to be willing to go outside of your own department frequently.

Link to comment
Share on other sites

WoW thank you so much for the introduction!

Actually I am not an Econ PhD yet, but applying for it right now, thus I want to bias my application to departments with exposure to this field. I can't say what specific topic I'm particularly interested in, as I haven't really entered the field. But I definitely want to dive into the area. What I'm expecting for myself (or imagination, illusion?) is that I will become a professional research economist in the industry (tech companies like you mentioned, or big financial institution) who has the ability to utilize ML/DL/AI techniques to solve economics problems, maybe estimations, maybe predictions, and etc.

Would you kindly provide advice on schools or specific faculty I should apply for? Thanks!

 

Susan Athey is a good start. OP, what exactly are you interested in with regards to ML in economics? You won't find economists developing new ML algorithms, but you will find a variety of applications.

 

(1) Some such as Chernozhukov at MIT are interested in areas like dimensionality reduction and using ML to estimate nuisance parameters. There is an area which probably begins with Chernozhukov at the top interns of rigor, and which focuses on deriving traditional econometric properties of estimators that are fully, or augmented by, ML-based. Simulation and performance are also considered.

 

(2) There is another area of researchers that are simply interested in what it would be like to replace traditional econometric estimation with ML. This is less likely to get you very far for a lot of reasons (chiefly: causality still rules the day in academic economics). See, for instance, Bajari et al. throwing 8 different ML methods at a traditional demand estimation problem and benchmarking out-of-sample performance against traditional IO models (there is some straw-manning happening here, but it's not that bad).

 

(3) Finally, there is another group of empirical researchers who don't "do ML" per se, but are interested in using ML for creative purposes such as unlocking new data sets, augmenting additional data sets, or benchmarking their econometric models in terms of out-of-sample predictive performance. No one here is involving ML at all in their statistical results of interest, they're just using it as a step in the process before the real econometrics are done.

 

These are, at least, the main categorizations as I see them right now in academia. I see (3) as being the most accessible and the easiest to get a foothold in, (1) as the most challenging and the least likely to be feasible unless you're in a top econometrics department, and (2) to be worth our while as a field to take a stab it--if nothing else out of curiosity, but very crude at this point and unlikely to look good as the core of a dissertation or JMP.

 

Economists working at places like Google and Amazon will be stepping even further outside of these boundaries. I'm personally working on combining aspects of (1) and (3) above. I'd be happy to point you to more resources if the area interests you. I'd say you need a strong foundation in econometrics, as well as coding chops and some computer science background. Beyond that, outside of the top 5, working in ML will require you to be creative with the resources at your disposal and you'll have to be willing to go outside of your own department frequently.

Link to comment
Share on other sites

Testing...

 

Susan Athey is a good start. OP, what exactly are you interested in with regards to ML in economics? You won't find economists developing new ML algorithms, but you will find a variety of applications.

 

(1) Some such as Chernozhukov at MIT are interested in areas like dimensionality reduction and using ML to estimate nuisance parameters. There is an area which probably begins with Chernozhukov at the top interns of rigor, and which focuses on deriving traditional econometric properties of estimators that are fully, or augmented by, ML-based. Simulation and performance are also considered.

 

(2) There is another area of researchers that are simply interested in what it would be like to replace traditional econometric estimation with ML. This is less likely to get you very far for a lot of reasons (chiefly: causality still rules the day in academic economics). See, for instance, Bajari et al. throwing 8 different ML methods at a traditional demand estimation problem and benchmarking out-of-sample performance against traditional IO models (there is some straw-manning happening here, but it's not that bad).

 

(3) Finally, there is another group of empirical researchers who don't "do ML" per se, but are interested in using ML for creative purposes such as unlocking new data sets, augmenting additional data sets, or benchmarking their econometric models in terms of out-of-sample predictive performance. No one here is involving ML at all in their statistical results of interest, they're just using it as a step in the process before the real econometrics are done.

 

These are, at least, the main categorizations as I see them right now in academia. I see (3) as being the most accessible and the easiest to get a foothold in, (1) as the most challenging and the least likely to be feasible unless you're in a top econometrics department, and (2) to be worth our while as a field to take a stab it--if nothing else out of curiosity, but very crude at this point and unlikely to look good as the core of a dissertation or JMP.

 

Economists working at places like Google and Amazon will be stepping even further outside of these boundaries. I'm personally working on combining aspects of (1) and (3) above. I'd be happy to point you to more resources if the area interests you. I'd say you need a strong foundation in econometrics, as well as coding chops and some computer science background. Beyond that, outside of the top 5, working in ML will require you to be creative with the resources at your disposal and you'll have to be willing to go outside of your own department frequently.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...