Interested in China's regional economic diversification ?
In this new paper in Regional Studies, we explore the role of relatedness, & high-speed rail, in China's regional diversification. The paper was led by lead @gaojian08
(1/n) 🧵
Paper: tandfonline.com/eprint/GRXBNTC…
We start by verifying that Chinese provinces are more likely to (i) enter related activities & (ii) enter activities present in geographic neighbors. These are classic findings in economic geography that we reproduce using Chinese enterprise data & firm financial data. (2/N)
Then, we compare these spillover channels. What matters more? Having related industries? Or a geographic neighbor that is already in that industry? We find that these two channels work as substitutes. (3/N)
Having many related industries, or many geographic neighbors where the industry is present, predicts entries. But having both is not more predictive than having one. As long as you have one source of spillovers (relatedness or geography) the other one is less relevant. (4/N)
Finally we look at the role of high-speed rail in accelerating regional spillovers. During the last decades, China's high-speed rail went through several acceleration campaigns, which should facilitate spillovers among nearby regions. (5/N)
These acceleration campaigns followed existing roads and rail lines (connecting cities which have been connected for thousands of years). So they were not designed to connect a particular industries (e.g. to connect advertisement companies in Shanghai & Wuhan) (6/N)
We find that regions connected by rail grew more similar in terms of their industrial structure. These findings, suggest that high speed rail contributes to spillovers among neighboring regions. (7/N). You can get the paper at: centerforcollectivelearning.org/papers
& tandfonline.com/eprint/GRXBNTC…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
How big are the digital exports of the U.S. compared to Europe?
Today, trade flows not only through container ports, but through routers. In this new @NatureComms paper, we introduce a method to estimate trade in digital products by combining machine learning methods with corporate revenue data.
Digital product exports, such as purchasing a video streaming subscription from a foreign website, are notoriously hard to estimate because tech firms own foreign subsidiaries. Moreover, existing service trade statistics lack the fine granularity needed to track digital products. /2
Two years ago, with @ViktorStojkoski, @philippmkoch, and @EvaColl8, we started an ambitious project with the goal of estimating digital product exports from corporate revenue data. /3
**New Paper**
Economic complexity methods are popular tools in industrial policy. Yet, despite their widespread adoption, these methods are sometimes misunderstood. In this new paper, I explain & explore the policy implications of economic complexity. /1
https://t.co/BQ6xLOA8cMbuff.ly/3YjBFbh
First, why are economic complexity methods misunderstood?
A key part of the confusion comes from the predictive nature of these methods. The concept of relatedness, for instance, anticipates the probability that a country or region will succeed at an activity./2
So the knee-jerk reaction that people get when they encounter these methods is to develop a strategy that recommends what the method predicts. These are the products/industries that would be easiest to achieve. But this line of thought is both wrong & incomplete./3
What is intelligence?
And how is it different from problem solving?
These questions are central in our current discussion on AI & were debated passionately this week at the Santa Fe Institute’s conference on collective intelligence.
But what did we learn?
🧵 .. 1/N
First, a disclaimer. In this thread I will focus on one idea, not all the ideas discussed in the conference, and will obviate other aspects of intelligence (eg multidimensional intelligence), not because these are not important, but because I want to communicate one point.
My focus will be on a distinction between intelligence & problem solving, because in my experience, when people are pushed to define intelligence on the fly they often gravitate towards a problem solving definitions of intelligence.
AI hype is on full swing, to a large extent, because of language models.
But as a writer, I am not totally convinced about the “productivity boosts.”
You see, writing fulfills a dual purpose. On the one hand, we write to communicate. But on the other hand…. /1
we write to clarify our own ideas.
We write to learn in ways that cannot be accomplished by reading.
A big part of what motivates a writer to work on a book is knowing that at the end of the journey I’ll be a different person.
You write not because you are an expert, but to become one.
Writing is sincere. It pushes you to encounter your own incompetence, repeatedly. And when your own words look stupid & your ideas malformed, you must either abandon them or refine them. In that process you learn.
One frequent criticism of economic complexity metrics is that some countries, such as Mexico or Slovakia, rank too high while others, like Australia and New Zealand, rank too low. But is this a problem with trade data or with the method?
Trade data does not perfectly reflect a country's capabilities because distance plays a role. Despite advances in communication & transportation technologies, it is still more convenient for a car manufacturer in Germany to work with suppliers in Czechia or Slovakia.