Tech Economics Newsletter
What's new from economist Andrew Chamberlain, Ph.D.
I hope this note finds you well. I'm here with my latest update from the world of tech, economics and data science:
Doing the Math on Gender Pay: We published a massive 70-page study digging into the gender pay gap in 8 countries last month. We got a lot of media coverage. Highlight: We looked at real-world job applications on Glassdoor for evidence of a gender "confidence gap," and found similar men and women mostly seek equal pay for equal work.
Teaching Deep Learning from Scratch: When teaching data science methods, it's nice to walk through simple toy models to help students build intuition. Here's a nice teaching tool walking through a simple "deep learning" or neural network model from scratch in Python. Try running it yourself -- here's my version of the code.
New Tech Economics Programs: The University of San Francisco has launched a new M.S. in Applied Economics program aimed at preparing students for data science and economics roles in tech, including data science skills, machine learning techniques, and the auction and market structure theory behind today's tech platforms. Similarly, the University of Oregon is launching a tech-oriented one-year program this fall.
Just Combine All the Models: Here's a nice new working paper from Susan Athey on using so-called "ensemble methods" combining predictions from many models, rather than searching for the one best model, when doing causal inference with panel data. Another nice synthesis of data science methods and economists' causal inference techniques.
Amazon's Top Secret Econ Squad: Here's a nice CNN profile of the large and influential economics team at Amazon. Managing to hire over 100 economists is an incredible accomplishment by Pat Bajari, and the success of that team is being increasingly noticed by other tech giants around the world.
Forget Deep Learning, Use Simple Models: Here's a nice talk making a strong case for using simple -- and even linear -- models for most data science problems. With so much discussion about deep learning and ML, it's easy to forget the power of fast, transparent, and easy-to-interpret linear models. I agree totally with this view.
Shorter, Smarter Surveys with Factorization: Here's a fun new working paper from Facebook's Sean Taylor on using matrix factorization and active learning to design shorter, more efficient online surveys. Most surveys are too long, and question responses are often correlated. We could measure underlying concepts with a lot fewer questions. This method does that, and fills in missing responses using probalistic matrix factorization.
Job Market Advice for Applied Econ Students: Here's a really nice Twitter thread asking for advice about how applied econ grad students can best position themselves for private sector tech and data science roles. Lots of good advice here -- including some from former Glassdoor researcher (and now Amazon economist) Ayal Chen-Zion.
Come Work With Me: My team is hiring for our annual summer data science and economics intern position at Glassdoor. Past candidates have gone on to economics roles at Amazon, Google, Moody's Analytics and more -- an awesome opportunity for an economics Ph.D. student looking for a future in tech.
Thanks for staying in touch -- I wish you a safe, happy and healthy spring.
Andrew Chamberlain, Ph.D.