More evidence of ChatGPT Disrupting Labor Markets

Evidence from Upwork spot markets

Today I’d like to break from my usual explainers of econometric methodology and introduce to you a new study that is making the rounds on ChatGPT and its impact on the freelance (“gig”) economy. The paper is titled “The Short-Term Effects of Generative Artificial Intelligence on Employment: Evidence from an Online Labor Market” and is written by professors Xiang Hui and Oren Reshef at Washington University and Luofeng Zhou at New York University. I will try to be efficient in my exposition by simply sharing the analysis they do, the methods they use, a brief discussion of their findings and my interpretation of the study’s implications. My commentary of the paper’s implications should be viewed as my discussion, not theirs, as the study is more positive than normative and I don’t mean to put words into their mouths.

So, to summarize at a high level what this papers is about, it is a study of the effect of ChatGPT-3 on freelance work at Upwork. Some of the pictures I will be showing you below, including two pictures from a couple of other studies I have been meaning to post about (and am grateful I now have the chance) will likely imprint in your mind the credible claim that labor markets are probably being affected by this new technology and that that effect may very well be permanent. And since Upwork gains as a portion of each transaction, Upwork itself too. Below I post a couple of teaser images describing some of those things so you can see what I will be discussing beyond the paywall of this substack.