TL;DR
- China has a critical role to play in addressing the 21st century’s most pressing catastrophic risks.
- In particular, the risk of a US-China war in the coming decades is real (Metaculus gives 50/50 odds of a conflict with >100 deaths by 2050, and there's perhaps a 15% chance of a war of the scale we're considering for this post). A conventional conflict could cost over 2 billion life years in the combatant countries even before taking into account nuclear escalation. Even less horrific wartime scenarios would reduce global GDP by double digits and plunge perhaps 5% of the world’s population back into extreme poverty.
- Analysis of modern China is very neglected relative to its scale. Only ~700 people in the United States conduct research on anything PRC-related for a living outside the US government. The US Intelligence Community does not have it covered, and a vanishing percentage of the 600 are oriented towards reducing catastrophic risk. Even more concerning is that the flow of researchers into the space has not increased even as US-China tensions have heightened (in fact, the early career talent pipeline is broken).
- Thankfully, there’s a lot of low-hanging fruit! Funding Ideas include
- Money to increase the pool of early career jobs, including direct funding for early career positions at think tanks
- A survey that explores the state of research and talent challenges, culminating in a roadmap to improve the quality of US policy debate around China more oriented around catastrophic risks.
- Epistemic tools like a giant centralized repository of Chinese government documents and translation algorithms tuned for government documents
- Philanthropically-funded research organizations built off CSET’s model that could inspire Congress to fund more impactful analysis
- Along with reducing the chances of a US-China war, an improved understanding of the PRC could prove invaluable for working on other catastrophic risks like AI safety and biosecurity.
A Word from the Authors
Jordan Schneider: I was the lead author of this report. For the past five years, I’ve run the ChinaTalk podcast and newsletter. In the process, I’ve interviewed more than 250 China-focused journalists, academics, and policymakers based outside of the PRC in recorded conversations for podcast episodes, and had at least 150 more casual conversations with folks in the field. On top of that, I've spent six years in the think tank field on China-focused work, spent two years in grad school in Beijing, and another two years in a macro-focused role at a hedge fund with significant investments in China. I have also interacted with over a hundred students who have reached out to me for career advice. The claims that follow build off those conversations, my work experience, and countless hours reading English-langauge China-focused academic literature as well as think tank and government reports.
I have scoped this paper to focus specifically on great power war in the next fifty years as it’s the topic where I have the deepest expertise. However, practically every major shortermist and longtermist risk area (biorisk, global economic growth prospects, space governance, geoengineering, AI safety) is intimately tied to China, plagued by the same analytical deficiencies in the English-speaking world, and could be served by interventions analogous to the ones proposed below.
Pradyumna Prasad: I was the supporting author. I run the Bretton Goods blog and podcast, helped write the sections on estimating the mortality and economic costs of a US China war, and graduated high school last year.
Summary
What's the problem?
In 1948, two years after Churchill made his Iron Curtain speech, the CIA had 12 Russian speakers on staff. In the following decades, philanthropists and the US government kicked off an intellectual mobilization that dramatically improved US policymaking, ultimately reducing the risk of direct US-USSR conflict.
With Speaker Pelosi's recent visit to Taiwan, US-China relations have reached their worst point since the Mao era. In such a geopolitical context, it may be surprising to non-specialists that China studies has yet to receive a any significant increase in investment. The current field is too small and too few of those with the requisite training and language skills are in a position to create the scale and quality of policy-relevant research requisite to the times. Without a relatively affordable investment by Congress and the philanthropic community to open up career paths for a new generation of China-adjacent analysts, the US is more likely to stumble into poorly optimized policy choices, needlessly exacerbating catastrophic risk. The same logic applies for pursuing suboptimal policies in how the US interacts with China with regards to great power competition, biorisk, AI safety, and space governance, among other topics.
The following focuses more explicitly on the US context as the American research and policymaking community oriented towards China is most likely to contribute to catastrophic risk and is the geography I’m most familiar with. That said, other countries whose actions will also influence the trajectory of future China-adjacent catastrophic risk (particularly India, but also Japan, Germany and the UK) suffer from even shorter benches of analysts.
Lastly, for the purposes of this exercise, I am discounting the possibility that US-affiliated EA money like OpenPhil or FTX could fund any research by mainland scholars aimed at directly shaping Chinese public opinion or influencing PRC policy.[1] [1]
What could a new philanthropist do?
A new philanthropist could aim broadly to invest in interventions that would reduce the risk of China-related catastrophe. This effort could start by commissioning a more systematic study of the institutional challenges facing the creation of high-quality and impactful China-adjacent analysis. Further interventions that I have a reasonable degree of confidence in strong ROI with include:
- Funding to open the professional aperture, allowing for more junior talent to flow into the field of policy-adjacent China research (via subsidized multiyear salaries for entry-level roles in government and think tanks alongside summer training programs) and buying talent out from academic and industry roles to focus on questions in the broader interest of improving policy to reduce catastrophic risk
- Funding for the creation of epistemic tools and training programs around analyzing contemporary China
- Commissioning research around funding for critical questions that would improve the quality of near-term decision points and illustrate the effectiveness of such funding for Congress to potentially take up the burden
Importance
The following section highlights great power conflict as a particularly salient risk that better understanding of China could potentially address. However, similar cases can be made around the returns from improving analysis to better inform engagement strategies with China around a wide range of priority EA issues, as well as potentially uncover new ones.
Great power conflicts aren’t great for humankind. The World Wars triggered the events that generated the largest losses of life in the 20th century. WWII alone directly led to the deaths of over 65 million people, more than the 40 million who died in wars over the subsequent fifty years. Far more suffered famine circumstances and long-term severe health effects from time spent in wartime. Lauren Gilbert has estimated that anywhere between 2x and 20x of DALYs come from beyond reported battlefield deaths in civil conflicts.[2] Given the severe and global nature of dislocations caused by great power conflict, I would venture the multiplier to be on the higher end of the spectrum.
Grave geopolitical dislocations brought on by great power war also open the aperture for horrific governance. The World Wars and Cold War directly caused the worst mass death events of the 20th century. In the 20th century, only six events (governance over the Congo Free State, the Russian Civil War, WWI, WWII, Stalin’s rule, and Mao’s rule) killed more than five million people.[3] Four of these are either great power conflicts or can be directly tied back to a great power conflict.
- Without WWI, the Russian Civil War which took nearly 10 million lives would not have developed as it did. Stalin would not have come to power, likely preventing the deaths of between 20 and 50 million people that were killed in his purges, the persecution of kulaks, the gulag system, and the policies that led to the Holodomor.
- Without WWII and Cold War tensions prompting Stalin to support him in the following civil war, Mao almost certainly would not have emerged in control of a unified China. Without Mao in the picture, the most likely alternative governments in power would be very hard pressed to enter timelines in which a number as large as the 40,000,000 who perished from purges, the Great Famine, the Cultural Revolution, and labor camps would have died thanks to government policy. Further, Mao achieved a singular achievement of screwing up economic policy so bad that over the course of 1949-1976, China barely grew.
Of the eighteen 20th century wars and atrocities with death tolls between 1 and 5 million people, by my shallow investigation, many were caused or severely exacerbated by great power conflict.[4]
Directly caused by a World War or Cold War dynamic:
- Chinese Civil War (1945)
- German Expulsions (1945)
- North Korea (1948)
- Korean War (1950)
- Vietnam War (1959)
- Cambodia (1975)
- Afghanistan (1979)
Reasonably exacerbated by a World War or Cold War dynamic:
- Sudanese Civil War (1983)
- Ethiopia (1962)
- Bangladesh (1971)
- East Pakistan (1971)
Less directly exacerbated by a World War or Cold War dynamic:
- Mexican Revolution (1910)
- Nigeria (1966)
- Burundi/Rwanda (1959)
- Chinese Civil War (1927)
- Congo War (1998)
Chances of Conflict
The base rate for great power wars was high before the 1800s. Industrialization probably led political elites to decrease their willingness to fight due to the increased costs of total war. In the last 20 years we hadn’t seen a single war between great powers (although some might argue that the conflict in Syria could count as one). The war in Ukraine is as close to a great power war as we’ve come in the 21st century.
![](https://39669.cdn.cke-cs.com/cgyAlfpLFBBiEjoXacnz/images/25d7ac3e5eb409866531e2779498763b1a3f47ba5b38aecf.png)
The sample size to select a base rate between the US and China might be Graham Allison’s book on the Thucydides Trap where he counts 12 out of the previous 16 emerging powers ending up fighting the reigning superpower. The odds of a war might be a little lower than what the pre-WWII numbers imply because mutually assured destruction (MAD) is still a credible deterrent. The chance of a war is probably not as high as 80%. However, taking the base rate in the nuclear age to be zero is also an error. The Cold War had four near nuclear misses (Berlin - 1948, Korea - 1952, Cuba - 1962 and ABLE ARCHER - 1983) which very easily could have caused a ton of harm.
China is ideologically motivated to take Taiwan and the US considers it to be a security priority not to have China take it over. Domestic public opinion in China is strongly in favor of it, and the US is moderately in favor of defending the island. Metaculus puts the odds of war at a 50% cumulative probability of a Great Power war by 2045 and a 75% by 2070. This is somewhat consistent with another Metaculus prediction of a 40% chance of China invading Taiwan by 2050. For a deeper dive on the topic, see Stephen Clare's 'How Likely is WWIII?'
Mortality in US-China Conflict Scenarios
As Andrew Krapevich recently wrote, “a modern great-power war that extended beyond 18 months would wreak destruction at a level and scope far greater than anything experienced in living memory.”
We modelled out the consequences of three scenarios here. BOTEC Model + Assorted Charts. See Appendix I: Mortality Calculus In Depth for a deeper discussion.
The subsequent numbers are very rough estimates for deaths in US-China wars in the following scenarios:
- Shorter war: A war is fought between the US and Chinese armed forces but is confined to Taiwan and East Asia. Neither side attacks civilians or the mainland of the other country but attacks enemy forces at sea and in Taiwan.
- Longer, limited war: There is a limited war between the US and China lasting more than 18 months but both countries choose to disengage when they see the costs of it. Neither side uses nuclear weapons but does attack the mainland of other country and there are civilian casualties
- Total conventional war: There is an 18+ month protracted war between the US and China lasting multiple years. Neither country chooses to disengage but neither escalates to the use of nuclear weapons. This would require total mobilization of both economies, leading to severe disruptions in international trade and economic growth, global famine conditions and a dramatic increase in extreme poverty.
Source of Loss | Short War | Limited War >18mo | Total War >18mo | Average |
Likelihood if conflict breaks out | 65% | 25% | 5% | – |
US Combat Deaths | 0.49 | 43.56 | 145.2 | 18.47 |
China Combat Deaths | 0.705 | 588 | 1764 | 235.66 |
Taiwan Combat Deaths | 8.924 | 49.45 | 98.9 | 23.11 |
Us injuries | 0.045 | 14.85 | 49.5 | 6.22 |
China Injuries | 0.0675 | 210 | 630 | 84.04 |
Taiwan Injuries | 1.035 | 17.25 | 34.5 | 6.71 |
Famine Deaths | 30.8 | 92.4 | 154 | 50.82 |
Extreme poverty | 61.6 | 154 | 462 | 101.64 |
Total | 103.6665 | 1169.51 | 3338.1 | 526.67 |
If the US and China were to come to arms, one might assume that armed conflict would resolve relatively quickly without much impact to welfare on a population-scale. However, Hal Brands in a recent report makes a convincing case that a drawn-out conflict in a Taiwan invasion scenario is a serious possibility.[5]
I would (with low confidence) assign the probability of a a shorter, limited war at 65%, a 18+ month limited war at 30%, and a 18+ month total war at 5%, though a prediction market could perhaps give a better estimation of a war’s severity.
There is also, of course, the risk of nuclear escalation in such a conflict. As Kit Harris from Longview Partners puts it, “A nuclear war would be a horror beyond comprehension. In the worst case, nuclear war could even lead to the permanent loss of humanity’s potential, putting it on a very short list of issues with potentially astronomical consequences. Nuclear weapons also interact with, and could exacerbate, other existential risks.” A US-China or China-India incident presents two of the more plausible scenarios for nuclear conflict in the 21st century.
Economic Costs of Conflict
The only extant somewhat rigorous projection of the economic impact of a US-China conflict was published in 2016 by RAND.
![](https://39669.cdn.cke-cs.com/cgyAlfpLFBBiEjoXacnz/images/3c43c88758b07149f5d15b3f12c2ffe124d9f79bf45fc0a7.jpeg)
However, this estimate seems very low. To quote communication with a former IC analyst: “ [The RAND] study only captures (some) direct effects, it doesn't factor in supply chains, and it applies fixed percentages for losses based on historical experiences.”
There are two places where the economic cost of a US-China war is most easy to predict. First would be the cost of disruptions to the semiconductor industry. Taiwan’s semiconductor ecosystem is likely to suffer irreperable harm, leaving a huge hole in the global microelectronics supply that wouldn’t be filled for a decade at least. Second would be general trade disruptions that can be somewhat backed out by looking at total trade volume and value add calculations. These impacts are by no means comprehensive and if anything a serious underrepresentation of the total amount of economic disruption.
The global GDP loss from a war would probably land somewhere between the impact on China and the US. The pain will be unevenly distributed, however, concentrating in particular on southeast Asian developing countries most dependent on trade with China, as US sanctions and wartime exigencies will prevent much of this trade from continuing. For a more detailed exploration of how we arrived at these very rough numbers, see our Economic Costs appendix.
Our Estimates (as a % of GDP)
China | Impact from semiconductors | General Trade Disruptions | GDP loss (relative to counterfactual) |
Short War | 15% | 7% | 12-17% |
Limited War | 15-20% | 8% | 18-23% |
Protracted War | 30-35% | 22% | 52-57% |
USA | GDP Impact from semiconductors | General Trade Disruptions | GDP loss (relative to counterfactual) |
Short War | 10-15% | 4% | 14-19% |
Limited War | 20-25% | 8 | 28-33% |
Protracted War | 25-35% | 10% | 35-45% |
World | Impact from semiconductors | General Trade Disruptions | GDP loss (relative to counterfactual) |
Short War | 12% | 5% | 17% |
Limited War | 20% | 6% | 26% |
Protracted War | 30% | 15% | 45% |
Neglectedness
Who is already working on it
The following estimate is very rough and some of the following numbers could very well be 2x off. With more funding to explore this question, it would be interesting to do some bibliometric work around China and CCP-related research (particularly in comparison to studies on the USSR in the Cold War era) as well as trace the number of think tankers and professors employed over time.
There are maybe two dozen fellows at the most influential Mass Avenue think tanks, another two dozen at American staff colleges, perhaps 250 academics, and a few dozen other Americans employed in China-focused roles in the private sector. This number pales in comparison with other directly policy-relevant research fields like macroeconomics.
There are perhaps fifteen senior fellows and another two dozen research assistants in their 20s with at least passable Mandarin in the most influential Mass Avenue think tanks (Brookings, CSIS, CNAS, AEI, CAP…). Federally Funded Research and Development Organizations (FFRDCs) like RAND and the Institute for Defense Analyses perhaps employ fifty more, while American military universities employ perhaps another hundred China-focused researchers.
The National Committee on US-China Relations’ Public Intellectuals Program (PIP) has trained cohorts of twenty US-based and China-focused academics every two years in how to make their work more policy relevant. This program, which has run for the past 15 years, can serve as a rough proxy for how many academics work at least somewhat aspires to make an impact in DC. Using the PIP program as a baseline, perhaps 250 academics’ work touches some China-related policy field. However, a far smaller segment of those researchers focuses on topics most germane to catastrophic China risk. On rough count, there are maybe three who focus on China + biotech (and none from a biorisk perspective), one who focuses half-time on China + geoengineering, five on China + space, five on China + semiconductor policy, a dozen on contemporary elite Party decision-making, 20 on the PLA in total and maybe five specifically on nuclear policy.
For an anecdotal sense of the gaps in scholarship on modern China, consider Xi Jinping. Anointed successor in 2007, for the past decade he has been arguably the second most important person on the planet after POTUS. However, it took until April of 2022 for the first Xi biography to be published (see this piece for some hypotheses as to why). In contrast, by my count, there have been well over a dozen serious biographies of Putin published in English, and more than a handful on Kim Jong Un.
To ballpark the number of other global academics outside the PRC contributing to China-relevant research that feeds up into DC thinking, perhaps the total number is around 700. With an average salary of $100,000, that would add up to $70m spent per year on China-related studies. Keep in mind though that almost all of this spending is very poorly optimized for research that would feed into strategies to reduce catastrophic risk.
In addition to the population identified above, another hundred Americans are employed in China-focused roles at private sector political risk firms, in corporate government relations positions and industry organizations. However, these positions’ work product is more tied to corporate rather than public interests and does not generally contribute to the quality of overall national understanding.
To put these numbers into context, consider macroeconomic policy research which feeds into the Federal Reserve’s decision-making. The Fed alone employs over 400 PhD economists, each of whom receives direct research assistant support and freedom from conforming to pressures associated with a tenure track appointment. The American Economics Association boasts 23,000 members, half of which seem to be practicing academics. The Fed defines the circle of expertise broadly, but for a conservative estimate one assumes that only half of the economists do significant work: that leaves nearly 6,000 economists in some way contributing to the discourse on Fed policy, to say nothing of the thousands of economists employed by financial institutions whose research is also often publicly available.[6]
Funders
The US Government
American intelligence agencies analyze China. The number of overall analysts, quality of research, and impact on policy are difficult to evaluate. Glimpses seen through Congressional testimony, declassified reports, and Wikileaks do not strike me as evincing a qualitatively deeper level of understanding from work from think tanks and academics. Former members of the IC have also contended to me that while classified work does substantially improve analysis out a 1-3 month time horizon, past 3-6 months the analytical edge gained from human and signals intelligence starts to rapidly degrade. Further, recent reporting like this November 2021 Bloomberg piece contends that “A lack of top-tier intelligence on Chinese President Xi Jinping’s inner circle is frustrating senior Biden administration officials struggling to get ahead of Beijing’s next steps, according to current and former officials who have reviewed the most sensitive U.S. intelligence reports.”
While intelligence analysts have the advantage of incorporating classified information into their thinking, bureaucratic exigencies commonly harm professional development and the final work product. For instance, rotation programs, narrow silos and demands for quick turn analysis on fast-moving events often hamper the deep time needed to develop expertise in the Chinese system. One former CIA analyst emphasized to me that the IC suffers most acutely from too many non-China experts working on China due to IC demand for China analysis growing faster than the IC’s ability to hire and train relevant talent. Other challenges include inefficient processes, cloistering of analysts within government, retention either through years-long rotations away from areas of specialization or attrition to the private sector, and managerial turnover. Multiple China-focused analysts who left the CIA and DIA have told me that only once they were outside were they able to do the deep, strategic research and writing they felt would be most impactful.
More broadly, policymakers are influenced by more than just classified US government reports. A 2014 study which interviewed over 200 senior American foreign policy practitioners (including participants all the way up to cabinet secretaries).
![](https://39669.cdn.cke-cs.com/cgyAlfpLFBBiEjoXacnz/images/4abff633f80b9565e9532a4074ac953ae0b62f268b1d4e25.png)
The study's authors argue the following:
Figures 16 and 17 show that the most important sources of information for policymakers are classified information and newspapers. This makes sense in terms of the unique resources inside government and also the limited time policymakers have to read outside materials. It is striking, however, that policymakers find newspapers as useful as classified information, lending more credence to the widely recognized–if seldom acknowledged –fact that most policy is made based upon open sources.
Newspaper reporters of course generally are not fed information and narratives from the intelligence community but rather from academics and think tankers.[2]
That all said, the federal government is a natural and promising source of sustained funding for this type of research, as China-focused analysis is fundamentally a public good. However, USG needs to be led to the water to drink. Fed with uninspiring government-produced, had officials not seen CSET in action, they may well have assumed that a higher level of quality was simply not possible to achieve. CSET’s reports have so impressed Congress that multiple staffers have told me their work has inspired proposed legislation like Sen. Bennet’s American Technology Leadership Act and Rep. Castro's Open Source Translation initiative. In order to fund further innovative research which will better inform policy around China and other EA priorities, philanthropies need to step up to run pilot research shops to demonstrate what’s possible to ultimately inspire the federal government to pick up the tab.[7]
Academia
Nathaniel Ahrens of the American Mandarin Society in a recent paper argued persuasively that the field's various institutions stuffer "organizational inadequacies" that limit talent's effectiveness. The handful of academics who have more freedom to pursue long-range research and China expertise are not incentivized to tackle policy-relevant questions. In the early years of the Cold War, regional specialists were eager to make themselves useful to policymakers. However, after the war in Vietnam, academia recoiled from such arrangements, and research interests and funding shifted away from disciplines like diplomatic history, military history, and the study of elite politics (for more see the book Know Your Enemy). As Ahrens writes, "academic institutions do not have policy advocacy as part of their DNA. In fact, in most of traditional academia, there is a cultural and institutional aversion to prescriptive policy work." What's worse, even for academics hoping to create research that will resonate in policy discussions, peer review and the demands of tenure shunt policy time to nights and weekends.
Corporations
Corporations occasionally fund China-focused research or work more generally aimed to shape US foreign policy as part of a government affairs or lobbying effort. A brief perusal of donor rolls at major think tanks shows that only defense contractors Lockheed and Northrop Grummond give consistently over $500,000 a year, which would raise obvious concerns that research could be bent in a direction not amenable to reducing the risk of great power war.[8] Most Government Affairs teams budget out support for think tank work on at the longest an annual basis for under $250,000 per project, hampering think tankers’ ability to invest in skills development (see CSET as a Model). This pool cannot be expected to a source of a sustained increase in marginal dollars going forward.
Foreign Governments
The Norwegian, Japanese, South Korean, Australian and Taiwanese governments all give away millions of dollars a year to support research in the US with the aim of influencing US policy. However, their money is also allocated on an annual basis in ticket sizes lower than $250,000 (often in the >$100k range). This pool comes with its conflict of interest concerns and cannot be expected to supply a sustained increase in marginal dollars going forward.
Foundations
Longstanding foundations including the Ford Foundation, Smith Richardson Foundation, and McCarthur Foundation among others do fund China-related analysis. However, this pool’s priorities do not align with directly with consensus EA community focus topics, and researchers suffer from small dollar sizes and short timelines (see CSET as a Model). This pool cannot be expected to supply a sustained increase in marginal dollars going forward, though outreach and growing salience of EA worldviews may help bend the dollars these foundations spend towards EA priorities that interact with China.
Substack Subscribers
Into the breach, independent publications like Sinocism, Neican, Discourse Power, Tracking People’s Daily and my newsletter ChinaTalk have attempted to fill the void. But unlike economic research or US political commentary, the market for contemporary China analysis is niche and comprises largely of government officials who have to spend out of their own pocket. To my knowledge, only Bill Bishop’s Sinocism is large enough to support one full-time position.
Tractability
Quantifying the impact of the marginal researcher on the chances of a great power war is not a straightforward exercise and overall is the most challenging link in making the case for increased study of China. The causal linkages between better understanding China’s intentions, capabilities and constraints, lobbying foreign governments to alter their policy, and that policy in turn lowering China-related risk is not a completely transparent process. However, I have a moderately high degree of confidence in the following assumptions.
Odds China significantly matters for catastrophic risk | 100% |
Odds western philanthropic funding for research can influence global policy towards China | 80-90% |
Odds that current resesarch on PRC realities and how to optimize engagement with China can be improved with more funding to support research | 75-90% |
Odds how the world engages with China could materially reduce China-related catastrophic risk | 40-60% |
Odds western philanthropic funding for research can directly influence PRC policy to materially reduce China-related catastrophic risk | 20-40% |
The previous hunches could be more refined applying to various specific catastrophic risks. For instance, perhaps the odds philanthropic funding for research could directly influence PRC policy around AI is lower and around biorisk is higher. Perhaps the 20-40% number on influencing PRC policy seems high, though Xi’s strong commitment to carbon reduction is evidence of the fact that modern China, while trending towards a more insulated system, is still very much influenced by global norms.
With respect to great power conflict, I’ll start by making the argument that analysis matters by resorting to 20th-century historical analogies. We’re fortunate that the N for great power war is relatively small. However, the fact that we haven’t had a ton of world wars makes the political science literature around trying to predict exactly what triggers global conflict pretty flimsy.
Historical examples abound of poor understanding of an adversary’s capabilities and intentions leading to decisions that increased the chance of catastrophic conflict and exacerbated the suffering caused due to war. For three 20th-century examples where more informed decision-making could have prevented the deaths of tens of millions:
- Had France, the US, and UK had a stronger conviction in the 1930s that Hitler would not be satisfied with piecemeal territorial expansion, they could collectively have taken a stronger stand, perhaps triggering a conflict before 1939 when the correlation of forces would have been much more dramatically in the Allies’ favor, or undermining elite political support for Hitler’s revanchism. (See: Hitler: 1936-45 Nemesis)
- Had the US better appreciated internal Japanese political dynamics, Army-Navy fissures, and tensions around the relationship between the military and Emperor, perhaps America could have crafted a more targeted export control and sanctions policy instead of one that ultimately triggered Japan’s bombing of Pearl Harbor. (see ChinaTalk Episode with Michael Barnhardt on his book Japan Prepares for Total War, Japan 1941)
- Had the US had a better understanding of the fundamentally nationalist motivations of the Ho Chih Minh and the tensions underlying the Vietnam-PRC relationship, it could have pursued a more conciliatory path in the 60s perhaps averting the regional conflagration that ultimately left ten million dead. (see Hanoi’s War)
In all three instances, there is a case to be made that more funding for independent study could have improved clarity and ultimately bubbled up to better policymaking.
Thankfully, investing in researchers is cheap! Capital costs are negligible, as financial support can flow directly into salaries. If the right talent is put in a position to do real support, a few people can have a disproportionate impact on the flow of government spending.
One study conservatively estimated a $220 return for every dollar invested in lobbying.[9] The Sunlight Foundation using a more liberal heuristic estimated $760 returned in subsidies or tax breaks for each dollar spent lobbying.[10] The impact of investment in think tankers researching and advocating around issues with a reasonable bipartisan consensus even higher leverage than lobbyists. Civic-minded researchers motivated with ‘cleaner’ money than corporations or foreign governments will meet far more receptive audiences in the Executive and Congressional branches than lobbyists pursuing largely transactional advocacy for corporate handouts. For a very tangible explanation of how a few policy entrepreneurs in think tanks can change policy, see this write-up of the creation of the U.S. International Development Finance Corporation.
What’s more, there’s a whole lot of federal budget to leverage.
![](https://39669.cdn.cke-cs.com/cgyAlfpLFBBiEjoXacnz/images/ec501b8f3952d62df25f7535c1ecaaf6673072ecd6a0ded3.png)
CSET as a Model
Open Philanthropy's support of the Center for Security and Emerging Technology could be seen as the minimum viable investment necessary to dramatically improve the quality of debate around a topic.
CSET research has transformed policy debates around high-skilled immigration, semiconductor export controls, military-civil fusion, and the role of government in managing AI’s impact (see here for their reading guide) by adding facts and analysis to debates bereft of them prior to CSET’s appearance on the scene. They also provide ‘reasoning transparency’ practically never seen in other think tank reports, expressing different degrees of confidence in claims and recommendations where most of its peers skate by on generalities and often-unsubstantiated assertions. Also, CSET is the closest thing America has to an open-source China research headquarters, having translated over 100 Chinese policy documents.
How CSET pull this off? Instead of traditional project-based small-dollar funding for a think tank, Open Philanthropy opted to give Jason Matheny and company $55m over five years in seed funding to advance a general research agenda. As someone who writes reports on budgets for a living, it’s clear that some of these CSET reports creep way beyond the $100-200k range you commonly see from policy shops and into the $750k+ stratosphere. CSET researchers have far more bandwidth to respond to inbound private requests from executive and congressional officials than your average think tanker too preoccupied with fundraising and executing her existing grants to do short or medium-turn pro-bono work. CSET also has the resources to buy commercial datasets, do bibliometric analysis, and run surveys far out of the price range of academics, government officials and other think tankers. It boasts a nine-person-strong data team that supports its 30+ deep analyst bench.
It only took two years to allow a former lawyer (admittedly with a BS in physics) in Saif Khan to blow past the knowledge frontier when it came to USG semiconductor policy. Remco Zwetsloot, who was pursuing a PhD in political science in an only tangentially related field, did the same with respect to high-skilled immigration policy. It’s not that Khan or Zwestloot are uniquely talented researchers or that they spotted the importance of their topics on the policy horizon earlier than others. Rather, CSET uniquely supplied the intellectual support and breathing space to these two, among a dozen others, to reach their potential.
CSET’s influence can be measured both in the impact its research has directly had as well as the careers it has set its staff up to pursue. Think tankers are also prototypical candidates for appointee positions in the executive branch. Jason Matheny’s experience running CSET made him a very competitive candidate for a senior tech policy position in the Biden administration. He leveraged that job even further into an opportunity to reorient $500m/yr of research towards more productive ends by taking a job heading up RAND. Other CSET employees also were able to leverage their demonstrated subject matter expertise into influential positions within the Biden White House and broader Administration.
CSET specializes in research adjacent to emerging technologies. There are, however, a wide variety of other China-adjacent policy areas (and non-China-adjacent policy areas for that matter) neglected relative to their global importance which could benefit from a replication of the CSET model. Investing in a deeper fundamental understanding of how the PRC system operates will pay dividends not only to issues surrounding great power conflict but more broadly in the panoply of EA-adjacent priorities where China will undoubtedly be a major player.
The Need to Rejuvenate China Studies
In 2015, former IC professional Peter Mattis wrote the following in the magazine National Interest.
February 2, 1977, the late Mike Oksenberg, a China expert then serving on the National Security Council staff, wrote a memo to his boss, National Security Advisor Zbigniew Brzezinski, expressing his dismay with the state of U.S. Government analysis on China. Although Oksenberg had “not been very impressed thus far by what [he had] seen,” he was admirably concerned about the future: “How do we cultivate talent so that 15–20 years from now, we will have a core group of 25–35 top-flight Chinese intelligence analysts in the then age bracket of 40–55. Everyone agrees with me that unless something is done, such a group will not exist.” Thirty-five years later, a senior U.S. academic on China would stand up at a Washington, DC conference and tell an audience of his peers in government and defense contractors that universities had failed to build expertise in the ivory tower and to produce sufficient numbers of properly trained analysts to support ongoing analysis of the Chinese military.
Mattis continues:
The training of PLA watchers has largely been an on-the-job endeavor. This means it is guided largely by tasks assigned and contracts won; it is neither systematic, nor strategic. In case anyone thinks government is substantially different, most intelligence training programs relate to improving analytic skills generally rather than developing country-specific or functional expertise. With the hiring binge of the 2000s, more and more government analysts learn narrower and narrower accounts. The situation inside and outside of government results in niche experts rather than people can pull it all together.
While Mattis refers here in particular to PLA studies, the same can be said for analysts of practically any other faucet of the Chinese system whose study could tie back to addressing catastrophic risk, including elite politics, foreign relations and technology policy of every ilk (see ‘Who is already working on it’). This paucity of good analysis is leaving a ton on the table for better research to more accurately tune policy around engaging with China to reduce catastrophic risk.
Since 2015, even as US-China relations have steadily degraded, very little has happened to improve this state of affairs. CCP-watching remains today a hollowed-out field. Few have the language skills, intestinal fortitude and career incentives necessary to devote years of reading open source documents to feel for how the Party and government operate. Those willing to put in this work, fearful of withering on the academic Galapogos of 'area studies', have been forced to cram their work into disciplines unsuited for deep study of a party state. With China closing and Hong Kong no longer the intellectual clearinghouse it once was, the quality of analysis is set to drop even further.
As Michigan Professor Yuen Yuen Ang in congressional testimony earlier this year characterized the gaps in contemporary China analysis in the following manner. She emphasizes that
While Chinese leaders, including President Xi, do formulate grand strategies, they are also human and fallible—often, they themselves are unsure how to reach a goal; they overreach and later retreat; or they completely fail to anticipate certain problems and reactions. In order to accurately assess China, it is necessary to take both ambition and fallibility into account.
In China policy outcomes are not only the result of leaders’ vision and elite politics; they are also shaped by policy communication and policy implementation, processes that involve a vast bureaucracy. In analyzing the sources of Chinese conduct, the latter has received little or no attention.
Specifically, on common prosperity, Xi has not yet found a way to resolve a fundamental conundrum: how to tame the excesses of capitalism without squashing entrepreneurial spirits. Precisely because he has no roadmap to follow, he has recently urged the officialdom to adapt and experiment.
Dr. Ang underlines there is more than enough open source material for such scholars to do meaningful work that would improve on the current level of understanding. If properly supported a new generation of analysts could give policymakers more understanding into CCP aims, likely responses to shifting economic and political environments, and institutional constraints unique to the PRC system.
On sources: Despite being a black box, there is a vast store of open information about the Chinese policy process, contained in speeches and directives. When read closely and in context, they can shed valuable light on leaders’ thinking and the internal workings of the Chinese bureaucracy.
She closed her critique with the following policy recommendation:
US government [should invest] in training a cohort of analysts who are able to read and interpret Chinese speeches and directives, which are an open source of information and available in vast quantity. Collecting these documents and translating them into English is tedious but not the hardest task—the hardest task lies in having the contextual knowledge to interpret them. This is long-term work that cannot be accomplished within few years, but it should begin as soon as possible, given that the most important bilateral relationship for the US in the coming decades will be with China.
As the US government hasn't done so, I'd argue that it behooves philanthropies to get the ball rolling.
Early Career Challenges
The field faces particularly acute early career issues. For starters, the number of American students studying abroad in China had been declining since 2010 before China's COVID Zero policy barred entrance for practically all foreign students. Now, the impossibility of studying in mainland China has dramatically reduced the appeal of spending the thousands of hours needed to learn Chinese to a professionally useful level. Given the staying power of China’s Covid Zero policy, such limitations do not seem to be going anywhere anytime soon.
![](https://39669.cdn.cke-cs.com/cgyAlfpLFBBiEjoXacnz/images/563192cc9f94b16a9d10f9e52d1eea53e6d47cd79d79bfbd.png)
Heritage speakers or advanced Mandarin students hoping us their language skills for the public interest have to land a very narrow set of jobs (I would estimate only there are only 25 full time openings per year in the think tank or think tank-adjacent (say Eurasia Group and Albright Stonebridge) sector for the under 26 set). Many positions for entry-level jobs require an expensive two-year masters degree. Upon graduating with MAs, most candidates will have to intern for six months or more (at, if they're lucky, minimum wage) before receiving a full-time offer with health care and benefits. Once a job offer is secured, salaries in the mid-five figures will struggle to make a dent in student debt, meaning a disproportionate number of folks who end up pursuing this career come from family wealth. Ironically, '内卷,' the 'rat race’ that Chinese youth bemoan in their professional lives, aptly describes America's China policy analysts.
Said one reviewer who spent almost a decade in junior and mid-level positions in a leading DC think tank:
In my experience, little...funding trickles down to providing for early career staff - to the point where using expanded funding to grow the number of early career roles is highly unlikely. Think tanks would rather spend that money on senior scholars or research trips, not adding new RAs. That feeds into a second point on think tank culture...
Second, while it varies between institutions, I think it's important to note that even with more unrestricted funding, multiple think tanks around town don't see it as part of their mission/culture to foster the next generation of experts. There are think tanks in town, like Brookings, that have deep challenges in this space. On the other hand, CSET, among other strengths you mention, seems pretty good at this. CNAS remains the gold standard for me. But I think it's worth grappling with the fact that there are places around town that - though they wouldn't say it out loud in public - wouldn't want a big infusion of money tightly tied to adding early career staff because senior scholars would see "dealing with" those people as impinging on their time. So there's a toxic culture dimension here that more money alone won't fix. But I do think that's addressable in your model by calling out some places suck at professional development and avoiding funding them for this purpose; that public pressure might push them to do better!
Initial Grant Ideas
Scoping the problem
China Studies needs a survey akin to what Alexey Guzey did for Life Sciences. A team of researchers should conduct an investigation exploring the broad contours of the China Studies ecosystem aiming to answer the following questions. This would cost maybe $1m for a really comprehensive job, though the value could be learned on narrower segments of the question from one person working full time on the question for only six months.
- What lessons can be learned from the ramp-up phase of Soviet Studies?
- What what the overall number of scholars/dollar amount poured into analyzing the USSR? Was it deemed sufficient at the time? What types of funding proved most and least impactful? Did it hit diminishing returns or what is possible that more eyeballs/dollars could have improved analysis?
- How many analysts currently work on China-related topics? How does this compare to other policy-relevant topics?
- What do policymakers see as the most severe analytical gaps on China?
- To what extent are those tractable?
- Analyst pipeline questions
- What is driving talent into the field?
- What is pushing talent out of the field?
- One potential methodology could be CSET’s use of survey data to explore talent flows in the AI field
In parallel, a team of researchers with fluency in EA priorities and some exposure to China could run down the list of salient EA topics and identify which face the most glaring gaps in analysis around China. This could be paired with a survey of the pools of analytic talent for said topic areas.
- China + reducing the risk of great power conflict
- Evaluate and right-size the pool of China nuclear policy analysts, PLA analysts, foreign policy analysts, elite politics analysts…
- China + AI safety
- Evaluate and right-size the pool of tech policy analysts, semiconductor analysts…
- China + biorisk
- China + space governance
- China + geoengineering
- China + demographic collapse
Commissioning specific pieces of research
Pressing questions could be addressed now to reduce risk if the current academic and think tank population was oriented more towards producing research to inform policy around catastrophic risk. Philanthropies could commission specific reports from think tanks or buy out academics from coursework to explore the following pressing topics:
- A detailed cost/benefit analysis of decoupling across economic and national security dimensions
- Scenario analysis of CCP stability exploring its impact on catastrophic risk
- Exploration of Chinese discourse around AI safety and biorisk
- A more detailed study of Appendix I and II explores the impacts of a US-China conflict
Reports of this nature that would move the needle on national discourse would cost at least $500k.
Content onramps and ongoing professional training
Undergraduate and graduate schools, to say nothing of online training resources, do not currently provide course offerings to help Chinese-speakers grapple with and contextualize primary sources. Some steps toward solutions include:
- Funding the creation and open-sourcing of online classes like Manoj Kewalramani’s ‘Watching China: Open Source Analysis of Chinese Politics & Policymaking’ that provide entry level and topic-specific handholding in using open source research to study the PRC by buying out practicing academics from teaching responsibilities.
- A program akin to the American Mandarin Society’s Syllabi Project but updated and incorporating the use of Chinese language source matieral
- China analysis/foreign policy summer camps for high schoolers and college students ala AI Summer Camps
Addressing Early Career Bottlenecks
As a quick fix, a philanthropy could subsidize slots at current think tanks for promising early career candidates with high levels of Mandarin fluency. It could also pair language training abroad with policy job offers (perhaps a summer interning at a think tank, then if performance is good a funded ‘gap year’ studying in Taiwan to polish language skills paired with a ‘return offer’). $10m over two years could double the number of young China analysts entering the field.
Organizations like the Atlas Fellowship could also consider building out China analysis-specific modules to inspire talented youth to consider careers around China-related catastrophic risks.
Tech Horizon Scanning
The west has only the vaguest idea of China’s progress on a number of critical emerging technologies with relevance for catastrophic risk. The horizon scanning cell proposed in this CSET report would provide researchers and policymakers far more visibility than is currently extant about a wide range of Chinese developments in topics like AI and biotechnology. This could start to make an impact for a few select industries at as little as $3m/yr.
Epistemic tools
There are a few million dollars in low hanging fruit around developing epistemic tools that would dramatically improve the quality of China-focused research across the board. In particular, the lack of a centralized database that captures the totality of Chinese government written output (work reports, newspapers, books…etc) makes research a real pain (my dream would be a cross-referencing Sefaria-style resource). Also, a machine learning translation algorithm tuned to translate Chinese government documents would also have a ton of leverage. I imagine these projects would be in the low single millions range.
Pilot ‘Think Tanks’ to attract USG funding
Philanthropies should look to employ the CSET as a Model framework to areas of particular neglected China-adjacent interest in order to in the next few years improve policymaking as well as encourage USG to pick up the tab going forward. Table stakes for seeding an impactful new institution is probably $3m/yr for at least 3 years.
Seeding expertise in government
A handful of mid-career foreign policy-focused fellowships exist (CFR’s IAF Fellowship), but they often bring in only two China-focused candidates a year into government. Given the importance of China to catastrophic risk, seeding EA-aware mid-career individuals into the executive and legislative branches could
Lobbying for improving US deterrence capability via acquisitions reform
What little credible research out there exists on defusing great power tension is the ability . The US acquisitions system is by consensus a broken beast completely (see the Acquisition Talk corpus or The Kill Chain for introductions). However, the defense industry’s $100m/year in lobbying makes even small changes hard to come by. Even devoting 10m/yr more enlightened acquisition policies could appreciably improve America’s deterrence capability, likely increasing the chances Chinese leadership give themselves of losing a war and in turn reducing the chances of them starting one.
Will this research have any impact?
One critique of funding policy research is that often the government will not revise policy in the face of new facts or changing reality. However, whereas research on highly partisan topics in the domestic sphere often amounts to little, the bipartisan consensus around China increases the chances that good work gets noticed and implemented. After all, the Chips and Science Act was the only piece of major legislation to receive substantial bipartisan support. In a likely divided Congress starting in 2023 and continuing for much of the next decade, China-focused legislation may well be the only policy topic that has a chance of moving. What’s more, funding this sort of research is a relatively cheap so the opportunity costs are rather low.
What if more research increases risk?
Some initial feedback I received probed on whether more research into China could accelerate competitive dynamics, raise the threshold for cooperation, and raise the risk of war. However, while mainstream media is baised towards these sorts of headlines, in my experience as much good analysis is devoted to popping hype balloons and more generally right-sizing risks. Take, for example, CSET’s research into quantifying the dollar amounts the PRC spends on AI. By putting in the work, CSET helped reset a misinformed talking point that made it into senior DoD speeches that China was spending 100x what it in fact likely was.
Of course, bad research is more likely to produce unintended consequences. But that’s all the more reason for EA-affiliated philanthropies to invest in bending the conversation away from hype and fear and towards quality epistemic practices to realistically assess China and its impact on the world.
Appendix I: Mortality Calculus in Depth
Sources of Death
Non-combatant country deaths
Food shortages are likely to be a major cause of death during a war between the US and China. The war would in all likelihood be concentrated in the Pacific. 60% of maritime trade runs through Asia and 30% flows through the South China Sea.[13]
and would disrupt a large amount of global shipping. It is likely that domestic ships would be requisitioned for military purposes, shipbuilding factories be used for military use, and maritime shipping insurance rates to skyrocket. Along with that, it is also likely that multiple countries will start to stockpile food in such a time which leads to the total quantity of food available for low income countries. The last time this happened was during WW2, when food shortages led to multiple famines causing millions of people to die.
For example, the Bengal Famine in 1942 led to 3% of India’s population dying of starvation. The Javanese famines of 1943-45 also led to a death rate of approximately the same amount. Over the course of WWII, the death rate among famine affected low income populations was 3% of the population.
Place | Years | Number of Deaths | % of population |
Bengal | 1943/44 | 2,100,000 | 3% |
Java | 1943-45 | 1,300,000 to 2,400,000 | 3-5% |
Vietnam | 1945 | 400,000 - 2,000,000 | 2%-11% |
Henan, China | 1942/3 | 2,000,000 | 5% |
Multiple reasons exist to update downwards from this number. The countries that suffered the most acute famine were controlled by colonial powers which were at best indifferent to their populations (and in China’s case, it had a Japanese invasion to contend with). That said, the governance challenges of keeping one’s populace fed through a third world war would be exceedingly high, and, as recently evidenced by Sri Lanka, many countries around the world will fail the test.
The main set of countries would be very low income countries with a GDP/capita of around $1000. The World Bank’s list here shows a population of 700 million for low income countries. I am scaling back the >3% estimates from the above table to a more reasonable .5% of the population dying from famine due to extreme food shortages in a total war, 0.3% in a limited war and 0.1% in a short war.
Extreme poverty
Extreme poverty will increase during a US-China war. There is also a substantial possibility of debt crises cascading on top of skyrocketing commodity and consumer goods prices. In COVID-19, an additional 1.4% of the world’s population were pushed into extreme poverty. For the years 2000-7, extreme poverty fell by about 1.25% every year but between 2007 and 2008, it fell by 0.7% which gives a very eyeballed rough estimate of a 0.5% increase in extreme poverty relative to the counterfactual.
A short US-China war would have as much of a shock as the global financial crisis with extreme poverty increasing by 0.4% relative to the counterfactual, in a limited war perhaps 1% and in total war this may rise to 3%.
Combatant country deaths
Estimating the number of deaths in the US, China and other combatants (likely Japan and possibly Australia and some European countries) is extremely difficult because we haven’t had a large-scale war with major powers since 1945. In WWII, Germany and the USSR lost almost a tenth of their population, while the US barely lost half a percent.
For a limited war where no civilians are explicitly targeted, I expect US troop losses to be in the range of 10,000 after adjusting upwards from the 2003 invasion of Iraq (~4k) and Afghanistan (~2k) because of the advanced nature of the enemy forces relative to both of them.
For a limited war, perhaps 0.3% of the US population (similar to WW2) dies because there is limited combat on US soil, and for total war, maybe 1% dies largely due to heavy Chinese missile strikes on US soil. These numbers are based on very shaky conjecture.
For China, I expect the numbers to be higher. Given lower levels of per capita income and lower nominal GDP, the Chinese government can probably mobilize lower amounts of their economy for military use giving the US an advantage in a protracted war. This would in turn lead to them having more casualties in the war. It is also possible that Chinese leaders are less constrained by public opinion and more willing to sacrifice people. Therefore, we are assigning a higher level of Chinese deaths relative to the US in every scenario. In a short war, 15,000 Chinese troops die, in a limited war, 1%, and for a total war 3%.
Taiwanese deaths are also an important factor. In every scenario they are likely to have more deaths than either country. In a short war 1% of Taiwanese die (similar to Ukraine’s ~0.9%), in a limited war it is 5% and in total war we may even approach 10%.
Non-fatal injuries
I have assumed as a simplifying assumption that nonfatal casualties are 150% of the number of deaths and each of them reduces life expectancy by about 10 years. I expect this to change if there was more work done on this, but I don’t expect it to alter the number of life years lost very dramatically (by 2x or more).
Total War Mortality Costs
In this model, there is a long US-China war with near-total mobilization but no use of nuclear weapons.
In this I assume
- 3% of the Chinese population dies
- 1% of the US population dies
- 10% of the Taiwanese population dies
- Injuries are 1.5x deaths (the higher end of WW2 numbers)
- Each non-fatal casualty loses ~10 years of life expectancy
- Famines kill about 0.5% of people in low income countries
- There is an increase in extreme poverty of ~10% (10% of the world’s population now earn less than $1.9/day now earn less than it) which leads to on average a reduction in life expectancy of ~2 years for each person.
Source of Loss | Millions of Life years Lost |
US Combat Deaths | 145.2 |
China Combat Deaths | 1764 |
Taiwan Combat Deaths | 98.9 |
US injuries | 49.5 |
China Injuries | 630 |
Taiwan Injuries | 34.5 |
Famine Deaths | 154 |
Extreme poverty | 462 |
Total | 3338.1 |
Extended Limited War Mortality Costs
In a limited US-China war where the militaries engage with each other, they hurt each other’s mainlands but the time frame is limited and both countries disengage when they see the potential for a larger war. This would see about .3% of the US population die and 1% of the Chinese one. Extreme poverty would increase because of the disruptions to trade and the loss of export markets for developing countries and the risk of a large-scale famine would increase. This may lead to 0.3% of the population in low-income economies dying due to food insecurity reasons and 1% of the world re-entering extreme poverty. If the latter leads to 2 years of reduced life expectancy, it would lead to 154 million life years lost.
Source of Loss | Millions of Life years Lost |
US Combat Deaths | 43.56 |
China Combat Deaths | 588 |
Taiwan Combat Deaths | 49.45 |
Us injuries | 14.85 |
China Injuries | 210 |
Taiwan Injuries | 17.25 |
Famine Deaths | 92.4 |
Extreme poverty | 154 |
Total | 1169.51 |
Short, Limited War Mortality Costs
This scenario defines a short war as less than one year with no land strikes hitting the US or Chinese mainland. Here I estimate that there would be limited deaths (~10,000 for American soldiers and~15,000 for Chinese soldiers), and global extreme poverty and food supply would be relatively moderately impacted. In this scenario, the extreme poverty rate increases by 0.4% (similar to what happened during the 2008-09 recession) and food insecurity kills 0.1% of the global low-income population.
Estimates for the life years lost are below:
Source of Loss | Millions of Life years Lost |
US Combat Deaths | 0.49 |
China Combat Deaths | 0.705 |
Taiwan Combat Deaths | 8.924 |
Us injuries | 0.045 |
China Injuries | 0.0675 |
Taiwan Injuries | 1.035 |
Famine Deaths | 30.8 |
Extreme poverty | 61.6 |
Total | 103.6665 |
Appendix II: Economic Costs of a US-China War
Low Confidence Estimates in Economic Impact of War
China | Impact from semiconductors | General Trade Disruptions | GDP loss (relative to counterfactual) |
Short War | 15% | 7% | 12 - 17% |
Limited War | 15-20% | 8% | 18-23% |
Protracted War | 30-35% | 22% | 52-57% |
USA | GDP Impact from semiconductors | General Trade Disruptions | GDP loss (relative to counterfactual) |
Short War | 10-15% | 4% | 14-19% |
Limited War | 20-25% | 8 | 28-33% |
Protracted War | 25-35% | 10% | 35-45% |
World | Impact from semiconductors | General Trade Disruptions | GDP loss (relative to counterfactual) |
Short War | 12% | 5% | 17% |
Limited War | 20% | 6% | 26% |
Protracted War | 30% | 15% | 45% |
Semiconductor Disruptions
TSMC makes over half of the world’s semiconductors. It gets two-thirds of contracts for semiconductor manufacturing and is one of the two companies that make 5nm transistors (the term somewhat funnily has no relation to the distance of 5nm) and will be for some time, the only company that can make 3nm chips. It might look as if the only impact of TSMC failing would be on Apple because it is its largest customer. But, transistors are used in several other applications like cars, data centers, missiles etc. TSMC’s next 5 customers after Apple are companies that make chips down the supply chain for other companies. It’s not just TSMC of course–many other smaller segments of the semiconductor supply chain are monopolized by Taiwanese production, particularly in tricky materials subsectors.
![](https://39669.cdn.cke-cs.com/cgyAlfpLFBBiEjoXacnz/images/609af755bd2eb6f65662a41636467dc19a383352c91fbd3c.png)
Should the Taiwanese chip industry be destroyed due to war or embargo, the consequences for the rest of the world would be dramatic. While the 9% of overall value capture may seem small, the capital investment required to replace that segment is easily in the trillions of dollars. Even once that money starts to be spent, multiple semiconductor industry analysts have told me that global production would fall to early 1990s levels for perhaps 15 years as the lag in production of fabs and tools, not to mention engineering know-how, would take decades to reaquire. In the meantime, global cloud compute would cease to grow, network coverage would slowly but surely start to degrade, and consumer good inflation would rise dramatically as no modern goods (PCs, cars, TVs, smartphones) could be produced. Given the importance of microelectronics to the global economy and the centrality of Taiwan to microelectronics, the disruption caused by such an eventuality would be hard to quantify.
To put Taiwan going dark into context, consider the most recent chip shortage. A Goldman Sachs report estimated that 169 industries were impacted by the chip shortage and there would be a reduction in GDP growth by 0.5 to 1%. This was because of a 20% shortage of semiconductors which led to slowing industrial production. But the impacts of this are non-linear, so merely extrapolating from this would be inappropriate. As Secretary Raimondo said in her blog post, American companies were holding 5 days worth of semiconductor inventory. But if this number went to 0, the impact could force entire industries to shut down. It also would be very hard to substitute for them because the entire world would be looking for semiconductors and it's likely that the regular trade routes in the Pacific would shut down.
Estimating the exact magnitude of those effects is very hard because there are several supply linkages. Michigan newspaper DBusiness says in an editorial that semiconductors form 12% of US GDP. I can't verify their numbers because they don't link to anything, but it's likely that the actual number is about that level. Statista also has a number of about 11% of GDP in semiconductor impact (but again I can’t verify the numbers). The other important point to note is that the 2021/22 situation was caused because of excess demand and slightly restricted supply. In a US China war, almost all the impact comes from supply going to zero from Taiwan which will have an impact on reducing quantities much larger than what happened in 2021.
If there were a protracted war, the effects of this would be reduced over time. American semiconductor companies would make things onshore. While it would lead to some of the quantity of semiconductor production being replaced, it would come at a higher cost.
China is completely dependent on the Taiwanese supply chain and will have an even harder time rebounding than the rest of the world due to sanctions, leading to massive economic impacts from the inability to secure microelectronics beyond smuggling channels.
General Trade Disruptions
As the RAND paper above notes, bilateral trade between combatants in a war usually drops by 90%. For the US it would mean that it would lose $140 billion in exports and China loses ~$500 billion in exports. For the US the loss of exports would be large but not devastating at .8% of GDP. China's nominal GDP is smaller and their export earnings to the US are 4% of nominal GDP.
In a limited war, it is likely that a large amount but not all exports will be stopped. Given that it would last for less than a year, losses should be limited to the loss in exports and the efficiency losses in trying to replace imports. For America, the losses due to export bans are minimal because it's a small percentage of GDP. The most serious cost would be that US consumers will now pay higher costs because they can't buy anything from China. Finding out the magnitude of this is difficult because in the recent past no countries have embargoed their largest trading partners, particularly in an era of globalization. However, we have some vague hints. For instance, these two papers show that the 2018 Trade War led to income losses for US consumers.
The effective tariff rate on Chinese goods was around 20% versus 3% for the rest of the goods. This led to a 32% decline of the imports of those goods (~$300 billion) which led to between $7.8 and $16.8 billion of income losses. A simple extrapolation would mean that there was a $96 billion decline in trade which led to a $12 billion decline in income. A $500 billion drop in trade would lead to at the bare minimum, a $40 billion drop in national income (~1.8% of GDP). It is also likely that because of the semiconductor shortage, American imports from Japan and South Korea would also be negatively impacted. For example, Japanese car production would be severely impacted leading to higher car prices and general machinery would be more expensive. Add to this the skyrocketing cost of shipping and overall trade flows would overall decline sharply.
Given that this would be an abrupt disruption in trade and that it would impact several intermediate goods, in the US the disruptions would be much larger than the 1.8% number. Adding up the effect on US exports to China (0.8% of GDP lost), it would mean that just by extrapolation, we can immediately forecast a 2.6% drop in GDP because of trade disruption alone. But even then, such an extrapolation is going to underestimate the costs of a US-China war. Entire industries would be shut down, which of course did not happen during the recent trade war. Shutdowns would stagger the broader economy, strain the financial system and lead to strained bank balance sheets across the world. Given that this is a back of the envelope calculation, a rough guess is that in a limited war, US GDP would decline ~4% from the trade disruptions, ~8% in a limited war and slightly higher in a long war (>10%) relative to the counterfactual of no war.
China—without embargo
In China too, the drop in trade would be disastrous for the Chinese economy. Over 4% of Chinese nominal GDP comes from export earnings and the multiplier effects of these are likely to be high. The effect of losing export earnings is going to be much higher on China than it is on the US. It is very likely that in the first year itself China experiences an over 6% of GDP drop because of the decline in export earnings alone and a further 2% of GDP from market access loss.
China—with embargo
In the more likely scenario where the US and its allies try to blockade China from exporting (in a protracted war), this could lead to a very deep recession in China. Chinese exports to the rest of the world are 18% of GDP and should those be blocked, it would be likely that there would be severe export losses. China’s export to GDP ratio is 18%, and if exports to the US are curtailed, they would be 14% of GDP. If the embargo reduces Chinese exports to the rest of the world by even half, that’s 7% of GDP lost. Counting the multiplier effects of this, it could mean that over 10% of GDP is lost (plus the initial 8% of loss due to the US exports).
The costs are higher for a protracted war because the US and its allies will have time to develop a more strengthened embargo. In this scenario perhaps three quarters of Chinese exports, amounting to 10.5% of GDP, are embargoed. Not only is the ~10.5% of GDP and any multiplier effects lost, but also the original 6% of GDP due to the US export loss. Taking the multiplier effect to be 1.5x, that’s 15.75% (~16%) of GDP lost plus the 6%.
Will government spending make up the difference from GDP losses?
This also has to be counterbalanced with the increased government spending in China to mobilize the economy for the war. It is likely that resources will be mobilized for the war effort. I don’t think these will meaningfully increase the GDP in the medium term. First, these are in the end unproductive investments. Iron used to make cars in which people ride is far more productive than the steel used to make guns to shoot other people. Aluminium used to make civilian planes is going to be of better use and going to increase economic growth far more than making fighters that will get shot down by/will shoot enemy planes. So, while the increase in government spending might kickstart an economy by increasing aggregate demand (like the US in the early 1940s) this is only true for very demand-constrained economies. On the flip side, it might lead to higher growth by increasing innovation, but estimating that beforehand is very difficult.
In sum, while GDP will go up relative to governments doing nothing, it will not relative to a no-war scenario. Along with that several non-GDP metrics would be severely compromised, (personal consumption in non-essential sectors would be down by quite a bit).
[1] Except for the Tsinghua-Carnegie Center, nearly all Chinese think tanks are government-affiliated. Most mainland academic institutions in this political climate would find it far too politically risky to take American money on a topic as sensitive as biorisk, nuclear policy or great power relations (imagine the inverse of the ‘Confucius Institute’ debate in a country where professors lose their jobs and get detained by police for criticizing government policy).
[2] https://forum.effectivealtruism.org/posts/ahranYvH9uum5dhtH/open-philanthropy-shallow-investigation-civil-conflict#Mortality_Cost
[3] https://necrometrics.com/20c5m.htm
[4] From https://necrometrics.com/all20c.htm
[5] From the Brands report:
First, countries fight long wars because they can. Since the Industrial Revolution and the birth of the modern nation-state, great powers have had the wherewithal—the economic resources, ability to raise taxes and field mass armies, and capacity to whip up popular fervor by appealing to nationalist impulses—to persist even after initial defeats. The events of the 19th century, writes Williamson Murray, revealed truths that have endured ever since: “That the new technological battlefield would take a heavy toll in lives, and that the capacity of the modern state to mobilize its human and industrial resources could feed that technological battlefield almost indefinitely.
Second, countries fight long wars because they feel they must. In great-power wars, the stakes are usually quite high, which gives countries a reason to keep fighting. In so-called hegemonic wars—clashes for geopolitical dominance between the world’s mightiest countries—the entire international order may be up for grabs…
Third, while democracies are no strangers to long wars, the price of defeat may seem even higher when autocratic regimes are involved. Dictators don’t fare well when they lose wars that they started. They are likely to be overthrown, killed, or exiled by their political enemies. This creates a strong incentive to “gamble for resurrection”—to double down, even on a losing cause—rather than accept a potentially fatal humiliation. The worse the personal consequences a leader faces for losing a war, the longer he or she will be willing to keep up the fight.
A US-China fight would likely fit the long-war pattern, for familiar reasons and more idiosyncratic ones. For one thing, it is far from clear that the early operations outlined by analysts on either side would prove decisive…
A US-China war would be likely to go long almost regardless of what happens at the beginning. If the United States did blunt China’s opening assault on Taiwan, Xi Jinping wouldn’t simply give up and send his forces back to their barracks. Admitting failure might lead to the overthrow of his regime or even his murder at the hands of his political rivals. It could lead to a declaration of Taiwanese independence with significant international support, leave a dominant America and its allies in position to dictate terms in the western Pacific, and severely damage his ambition of making China the world’s leading power. In other words, whatever goals Xi sought to advance by starting the war would be gravely endangered by losing it quickly. So he might keep fighting even if the struggle was not going well. “A failed Taiwan landing would not end the war,” writes Defense Intelligence Agency analyst Lonnie Henley. “China would continue the conflict by whatever means available.”
[6] The Fed defines the circle of relevant research as broader than strictly macro to include “behavioral economics; econometrics and data science; economic measurement; financial institutions and markets; financial stability and macroprudential policy; foreign economies; industrial organization; international trade and finance; labor; macroeconomics; microeconomics; monetary economics; payment systems; and/or public economics.” Perhaps the right analogy for China policy would also include adjacent fields like international relations scholars and historians of US and East Asian foreign policy, which would maybe double the number to 1200.
[7] While the government already does a decent job supporting language learning and research through programs like Fulbright and the Boren and FLAS, these programs with the qualified exception of the Boren do not open up career pathways. Although this is Boren’s ostensible aim, anecdotally, many Boren scholars have told me that in fact, they were not able to leverage their language skills in their placements.
[8] CSIS: https://www.csis.org/corporation-and-trade-association-donors
CNAS: https://www.cnas.org/support-cnas/cnas-supporters
[9] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1375082
[10] https://sunlightfoundation.com/2014/11/17/fixed-fortunes-biggest-corporate-political-interests-spend-billions-get-trillions/
[11] Data from https://opendoorsdata.org/fact_sheets/china. Taiwan hosts around 1000 American students a year, though that number has also been trending downwards as well even after the PRC closed its doors to students.
[13] https://chinapower.csis.org/much-trade-transits-south-china-sea/#:~:text=The%20United%20Nations%20Conference%20on,one%2Dthird%20of%20global%20shipping.
[14] https://voxeu.org/article/famines-wwii
[15] https://web-assets.bcg.com/9d/64/367c63094411b6e9e1407bec0dcc/bcgxsia-strengthening-the-global-semiconductor-value-chain-april-2021.pdf
- ^
Except for the Tsinghua-Carnegie Center, nearly all Chinese think tanks are government-affiliated. Most mainland academic institutions in this political climate would find it far too politically risky to take American money on a topic as sensitive as biorisk, nuclear policy or great power relations (imagine the inverse of the ‘Confucius Institute’ debate in a country where professors lose their jobs and get detained by police for criticizing government policy).
- ^
See https://scholars-stage.org/american-policy-makers-do-not-read-books/ for more on the paper
Thanks Jordan and Pradyu for this very thorough overview! Also still digesting much of the content, but wanted to note that we (Oxford China Policy Lab) are in the process of drafting a research agenda on intersectional risks between US-China GPC and AI development/governance (project supported by BERI). We will likely expand this work into a longtermist research agenda focused more exclusively on other risks related to US-China GPC and hope to hire a research assistant for this when the time comes. In the meantime, we'd love to be in touch with anyone else focusing on these issues. Cheers!
I'm also very interested in this topic; I taught for Chinese University of Hong Kong- Shenzhen (CUHK-SZ) for a year, including some EA content in my decision-making and evolutionary psychology classes.
I agree with Jordan Schneider that US, UK, and Euro researchers -- especially those considering global catastrophic risks -- need to develop a much more accurate, up-to-date, and sophisticated view of China.
And of course, one good way to do that is to listen to more of Jordan's excellent China Talk podcasts!
haha Kayla, Scott promised me he'd introduce us two weeks ago! v excited to chat
Thanks for this. There's a lot to digest and I'll come back in a day or two with a longer comment. But I wanted to quickly jump in and suggest you revise the second bullet point in your summary. It reads:
This is not true. The Metaculus question you're referring to is this one, in which the median forecaster prediction for the next great power war is 2045. However this is not specifically about a US-China war. Rather, this includes any war between two countries which are in the top 10 by military spending according to SIPRI's index. That includes the US and China, but, crucially, Russia, Germany, India, Japan, and South Korea (among others) as well.
Given the number of potential combatant-pairs in that set, some of which are probably about as likely to go to war as the US and China, I think your bullet point as written is very misleading. The odds of a serious US-China war by 2050 are closer to 10-20% IMO (a more specific Metaculus question has it at 16% by 2035).
There's also a Metaculus question about the chance of >100 deaths in US-China conflicts before 2050, with the community prediction at 53%. But 100 deaths is well below the threshold for a war and way, way below even your "shorter, limited" war scenario (which assumes 25,000 combined US-China deaths, plus >200,000 Taiwanese deaths). Because war deaths follow a heavy-tailed distribution, I think a lot of the conflict probability mass lies between 100 and 25,000 deaths.
I think a 10-20% chance of war is probably still high enough for many of your arguments to go through. But I think starting your piece off with such a misrepresentation does it a disservice. I recommend you edit that bullet point.
(Like I said - overall I'm really happy to see this piece and will come back with a more substantial comment in a day or two!)
Thanks for this Stephen. We wrote the TLDR at the last minute and switched out the 2070 for this one. Will adjust the draft accordingly to reflect this. However I gotta say the fact that most of the China war metaculus markets barely moved in the leadup to the Pelosi visit did make me downweight my confidence that the predictions in these markets mean much.
Metaculus performed decently well anticipating the Russia-Ukraine conflict, so I do think they can provide information about conflicts
still very excited to read your longer thoughts!!
;(
just came across your series--should've just outsourced my estimating to you you did a good job in those posts
Again, thanks both for this! Really appreciated you taking the time to collect these data and put your thoughts down in writing. Some more thoughts after coming back to this recently:
Adding some commentary from folks who aren't on EA Forum but gave me their permission to post:
Dr. Lin Zhang (https://cola.unh.edu/person/lin-zhang)
"I agree with lots of the ideas you put forward in the proposal. From my perspective as a communication scholar working in a public university who has been studying and later working in the US since 2007, I see a drastic change since the latter years of 2010s in terms of interest in and funding for China studies in academia...though we hear more about China in the news. I am quite concerned by the representation of China in mainstream media.
I agree with you that many academics are perhaps not policy-oriented with their research, but I believe academic research has its own merits. They often more historicized, contextualized, and take a long-term view. I think the key here is to establish more channels to translate academic research to serve policy goals. Yours and several other media outlets like Sinica Podcast does a good job giving academic research some more public exposure and get academics to become more policy-minded.
Also, from my own perspective as a China-born international scholar who has been studying Chinese international studies and more recently US-based ethnic Chinese scholars in STEM, I think there's also a concern about how inclusive the US will be to ethnic Chinese scholars given the deteriorating US-CHina relations. Personally, I think these scholars play an important role of bridging the two societies. They could offer more policy support if given a better platform in the US.
Another big issue here is how to fill the vacuum left open by the Confucius society. For public universities like ours, Chinese language teaching used to be heavily reliant on funding from CI, now they are gone from most American campuses, who will fund Chinese language studies?
CI also used to organize a lot of cultural events and to bring in speakers for academic talks, now we no longer have that budget."
Incredible amount of research and insights - still reading as well - you might as well write a book :)
Posting some anonymous feedback I received from someone who has spent a number of years in a Mass Ave think tank:
"First, one place where I think you can further reinforce the argument is doubling down that corporate support for think tanks is not going to solve this problem. In my experience, little of that funding trickles down to providing for early career staff - to the point where using expanded funding to grow the number of early career roles is highly unlikely. Think tanks would rather spend that money on senior scholars or research trips, not adding new RAs. That feeds into a second point on think tank culture...
Second, while it varies between institutions, I think it's important to note that even with more unrestricted funding, multiple think tanks around town don't see it as part of their mission/culture to foster the next generation of experts. There are think tanks in town, like Brookings, that have deep challenges in this space. On the other hand, CSET, among other strengths you mention, seems pretty good at this. CNAS remains the gold standard for me. But I think it's worth grappling with the fact that there are places around town that - though they wouldn't say it out loud in public - wouldn't want a big infusion of money tightly tied to adding early career staff because senior scholars would see "dealing with" those people as impinging on their time. So there's a toxic culture dimension here that more money alone won't fix. But I do think that's addressable in your model by calling out some places suck at professional development and avoiding funding them for this purpose; that public pressure might push them to do better!
Third, both in terms of great power conflict, but also even more so around competitiveness, it seems worth discussing why corporate won't fill some of the gap in its own for-profit need rather than through non-profit philanthropy. You mention Eurasia Group and ASG in a line, but I could see some thinking as corporations need to manage their interactions with an increasingly fraught China market, won't they have to hire knowledgeable China experts and won't that be a demand signal? I think you and I would agree the answer is (a) possibly no; and (b) even if yes, it wouldn't be a sufficient number or make sure good public policy is made by private-trained people (not that they can't contribute, of course; we just benefit most from a healthy mix). So that just suggests 2-3 sentences defanging a counter-argument that would further strengthen yours.
Fourth, your highlighting early career bottlenecks make a lot of sense to me. Just one additional point you could make on that is those bottlenecks also spur a lot of early career drag. It's not just that the limited numbers of opportunities box people out, even the people who make it through the eye of the needle are hampered in their productivity because they have to spend an inordinate amount of time fighting to get through the bottleneck and preparing for the next one. Early career opportunities are often time bound, so even when successful they immediately have to invest time in finding the next fellowship, job hunting, etc. Even the fortunate ones are not working/developing at anywhere near their capabilities because of this persistent time suck.
Finally, on scaling the CSET model, I think it's right in substance. More CSETs are what the policy community needs to inform their work. My question about whether it works in practice comes from my read of the think tank world as a traditional biological ecosystem. To compete and survive, each organism finds its specific niche based on a combination of (a) the substance they focus on; (b) the types of products they produce; and (c) any political leanings. Having many of a single type doesn't quite work because they crowd each other out for attention and struggle for funding. Will philanthropic funders be OK with their investments being one of several good think tanks doing work on this, or will there be pressure to focus on the best in a way that drives competitors out of the market? To put it another way, does the think tank market make sense to build several CSETs or to super-size the existing CSET to do more? I think from a good policy perspective, multiple CSETs providing different analytical inputs is better; what I don't know is if funders will see that value as sufficient to long-term support something seen as "good but not the clear best."
Just some thoughts as you continue on next steps. But none of it takes away from the fact that I think the piece stands extremely well as-is in terms of being correct in its analysis, making a necessary point, and doing it well. It's just fingers crossed funders see the wisdom!"
Another comment from a longtime China-focused CIA analyst:
"I read your pitch and very much enjoyed it, thanks for sending it! Here are some scattered thoughts and reactions:
My general view is that two things are true:
I know little about the effective altruism approach, but I think I'm guided by a different set of criteria. The question for a philanthropist might be "where can I best spend my money amid the vast universe of worthy causes" and for me it's more like, "should the US reallocate national resources from existing national security investments to this?" And I think the answer to the latter is unambiguously yes. Another CSET probably costs like 1 training sortie for a fighter jet...
I personally would ascribe drastically lower probabilities than you that philanthropic funding can influence PRC or US policy to materially reduce China-related risk. (Although again, I think it's still absolutely worth bolstering our knowledge base and talent pipeline!) I'm skeptical because I think the primary drivers of policy decisions are other things like material factors, ideological dispositions (derived from an emotional base that is hard to move with fact-finding and analysis), and self-interest on the part of organizations and states.
I had a few other random specific comments, not really disagreeing with what you've written but just thoughts that your piece sparked:
Again I really enjoyed your pitch, it was very thought-provoking, and I hope your suggestions are getting audience and traction!"
Thanks for posting these, great to read these perspectives!
Thanks for making this case. More investment in China studies seems straightforwardly valuable.
(Still reading.)
thanks for reading!