Tech Economics Newsletter
What's new from economist Andrew Chamberlain, Ph.D.
Spring is here, and I'm back with my latest update from the world of tech, economics and data science:
Making My Own News: I've been busy with media this month. I'll share two favorites. Here's my take on America's slow wage growth on NPR's "All Things Considered." And here's me complaining to Reuters about having to learn a new programming language every 18 months.
One A/B Test to Rule Them All: Economist David Reiley is a pioneer for field experiments in tech -- and a really nice guy. He runs the research group at Pandora and just published an epic paper -- with Jason Huang and Nick Riabov -- on a long-running experiment affecting 35 million Pandora customers. They show the causal impact of advertising on time spent listening to music is 3-4x smaller than observational estimates. If you're looking for one paper showing how to use causal inference to do awesome things in tech, this is it.
How Many Data Engineers Per Data Scientist? Many companies think they need more data scientists. Often what they really need are data engineers. What's the right ratio between data scientists and data engineers on your team, and when do companies need each? Here's a terrific explainer from Jesse Anderson.
Construction Is the New Tech: Tech and data science have transformed finance, e-commerce, health care and advertising. Now they're also reaching blue collar fields in construction -- and changing how we build cities. Here's a fascinating webinar on how tech is helping drive productivity gains in today's construction industry.
America's Hottest College Major: The University of California, San Diego rolled out a new data science undergraduate major in Fall 2017. Already, booming enrollment has prompted the university to cap the program starting this fall. Get ready for an avalanche of young talent moving into data science soon.
If Your Data Are Bad, Your Machine Learning Is Useless: A gentle reminder from Harvard Business Review. Most companies today think they need deep learning, collaborative filtering and ML platforms. But the first step is clean data at scale. Most data science work is 90 percent cleaning messy data. If you want to run scalable ML models, you've got to deal with that problem first.
How to Run a Data Science Team Like Zillow: Looking for practical advice on how to build a world-class data science team? Here's a handy summary of the tool kit currently in use at Zillow, as presented at NABE's 2017 tech economics conference in Seattle.
Companies Are a Garden, Not a Juice Machine: We turned over Glassdoor's database of several million company reviews to a research team at the University of East Anglia recently. The result? A new working paper linking employee ratings and financial performance in the UK. Here's my write up of the paper.
Thanks again for staying in touch -- and best of luck in the coming month.
Andrew Chamberlain, Ph.D.