China’s next policy change: Not what President Trump hopes for



Weifeng Zhong, Julian Tszkin Chan | American Enterprise Institute

As the US-China trade fight drags on, which side will back down first? Many, including President Trump, seem hopeful China will soon blink. But an AI program we recently developed, the Policy Change Index (PCI) for China, suggests otherwise.

The PCI is a machine learning algorithm that predicts policy shifts in China by “reading” its official newspaper, the People’s Daily. The idea is straightforward: The government often alters discourse in state-run media before changing course on policies.

The front page of the Communist Party’s flagship newspaper the People’s Daily (C) and other newspapers are seen one day after the unveiling of the new Politburo Standing Committee in Beijing, China, in this photo illustration taken October 26, 2017. REUTERS/Thomas Peter

According to this algorithm, China will take an even more aggressive approach on the global stage, especially in high-tech and trade-related areas.

Figure 1 shows the quarterly PCI for China, which runs through the end of 2018. The spikes in the index have “predicted” many historical policy moves that follow. In the first quarter of 2018, the index shows an uptick, which does suggest an upcoming change in Beijing’s policies. However, the shift is anything but what President Trump hopes for.

Figure 1: PCI-China shows a sign of change in 2018 Q1

What will the change be? The answer lies in how the index is constructed.

As explained in our research paper, the program reverse-engineers the People’s Daily editor’s mind by learning how to correctly pick out articles that ended up on the front page — the most prominent space in the tabloid. The algorithm trains itself over a five-year period and, subsequently, tests its “knowledge” on the next quarter. (Spoiler alert: It may not hold up well.) When the index spikes in the first quarter of 2018, it means the program suddenly makes a lot of mistakes in that quarter, as compared to the five-year training period — hence, signaling a major change.

When the algorithm blunders, there must be many articles that the machine deems unimportant based on past experience but that, surprisingly, land on the front page. It is these mistaken articles that suggest the future direction in which China’s policies are headed.

Figure 2 delineates the major policy issues underlying the machine’s mistakes in the first quarter of 2018. The bars show the algorithm’s error rates by topic, for the first quarter of 2018 (red) and the five years leading up to it (blue). The algorithm blunders in a certain issue when the red bar is much higher than the blue bar.

Figure 2: Blunders of the PCI algorithm by topic

(Note: The error rate metric is the false omission rate, the fraction of false negatives in all predicted negatives. See our research paper for more details.)

Several patterns in the chart are obvious. First, the Communist party is strengthening its authority internally, including ratcheting up President Xi’s trademark — the anti-corruption campaign. This is hardly a surprise given President Xi’s apparent ambition to stay in power for a substantial period of time.

Second, China is playing up the nationalism card and projecting a stronger image on the global stage. It emphasizes the importance of science, technology, and their priority in China’s increasing trade with and investment in the rest of the world. There’s no question this has become a thorny issue to the US, but the PCI algorithm now suggests a turn — for the worse.

Third, President Xi’s internal and external aggression can be explained by some policies that keep him rather popular in China. Besides the anti-graft efforts, China is enhancing a set of populist measures: the support for the agricultural sector, the fight against poverty, and the policies that help migrant workers. Such a wide popularity also implies, should China back down in the trade fight, a daunting political cost on President Xi himself.

The trade conflict has taken a toll on the Chinese economy. But it does so too on the US economy. The fact that it is a game of chicken does not imply any side is necessarily going to blink first. For China, “buying (more of) America” is easy. But satisfying the more pressing demands from the US — downsizing the state sector, protecting intellectual property rights, and abandoning its industrial policies — would require fundamental changes to the Chinese economy that typically call for justifications from the state-run media in advance. If President Xi is contemplating any of these changes, the People’s Daily is not showing any trace of it.

The upcoming visit to Washington by President Xi’s top aide Liu He will surely offer some clues as to how the trade talks will play out. But for those who think Mr. Liu will bring a major concession to the table, here’s our algorithm’s advice: Curb your enthusiasm.

To view the original paper published by the American Enterprise Institute, please click here.

Copyright © American Enterprise Institute. All rights reserved.