#293 The Past is an Incomplete GuideAmerican Tariffs and the Lessons from History and a Discussion on a new AI Governance Paper
Global Policy Watch #1: Back to the BasicsInsights on global issues relevant to India— RSJA couple of days back, I nearly spilled my coffee as I read this piece of advice from our Commerce and Industry minister to the Export Promotion Councils (EPCs) that were meeting him:
Hmm. Let me swallow the irony pill and look at the data here. In 1991, before the famous Manmohan Singh budget dismantled the duty regime, the average import tariff was about 120 per cent on a simple average basis. Over two decades, successive governments followed the path of lowering trade barriers and greater integration with the global trade system, bringing this down to about 13 per cent by 2014. In the last decade, we have had over 3000+ instances of tariff increases across goods as part of the Aatmanirbhar Bharat agenda, which has reversed this trend. India’s average tariff stands at over 17 per cent now, primarily driven by a protectionist mindset and the industrial policy followed by the Commerce Ministry. It is a bit rich then for the minister to ask EPCs to “come out of their protectionist mindset”. Richer still when you consider back in June 2020, at the peak of the first COVID-19 lockdown, when domestic demand had collapsed, leading to a collapse in imports, it was the same minister who had rejoiced over the goods trade surplus. All we need is for our largest export partner to turn into a tariff bully so that we can see the virtues of low tariffs and free trade. The global simple average tariff rate is about 5 per cent (compared to 17 per cent of India), and almost every large export partner of India has a comparable rate that’s less than half of India. As a long-term critic of tariffs, I’m happy that, acting on the coercive threat of reciprocal tariffs from the Trump administration, we are starting to bring tariff rates down and sending a message to domestic manufacturers to shed a protective mindset. This worked for the Indian industry and the economy through much of the post-liberalisation era, and there’s no reason why it shouldn’t work again. Tariffs make no sense - they are a drag on trade, a tax on consumers, they distort market outcomes and a deadweight loss for the society. India has got a Trump-sized massive excuse to reduce tariffs and make it palatable to its domestic constituents. We should welcome this collateral benefit of Trump’s deranged love for tariffs on our trade policy. Smoot-Hawley Tariff Act and lessons from historyThere’s been a lot of discussion about the Smoot-Hawley Tariff Act in light of Trump’s tariff frenzy and the risk of a reciprocal tariff war. In 1930, the Smoot-Hawley Act raised tariffs on agriculture imports and about 20,000 other goods. This was on the back of Hoover’s campaign promise in late 1928 to raise tariffs, especially on agricultural products. Following WW1, European agriculture took time to recover but it meant a boom time for the American farmers who borrowed to increase capacity and overproduced. Once the European farmers also recovered, there was further oversupply, leading to a sharp fall in food prices worldwide. The heavily indebted American farmers found it difficult to pay off their loans, and it was during this moment of crisis Hoover proposed his plan to protect US farmers. Soon, you had other interest groups asking for similar favours. Thus came the Smoot-Harley Act with its 20,000 plus goods in its ambit. The Act raised tariffs by about 20 per cent across the board, and soon, the other countries retaliated. The immediate impact was seen in the equity markets. Foreigners withdrew capital, imports became expensive, leading to inflation, and other countries retaliated with tariffs on US goods and agriculture, and this aggravated the economic depression that had set in. However, the most significant impact was on global trade, which declined by about 65 per cent. Most countries took the Act as a signal that the global trade order with its implicit agreements was broken and imposed tariffs not just on US goods but on other countries. It took years to come out of the impact of the Act, with a gradual lowering of the tariffs and multiple bilateral trade agreements leading to an increase in global trade by the end of the decade. The Trump administration isn’t reading any history (or any economics). Here’s an exchange:
I think it is insulting to the discipline of economics when the Press Secretary makes such arguments in favour of tariffs. But then ignoring sane advice from economists against tariffs and then vilifying them is part of the playbook of every administration across the world that has gone down the path of protectionism. Even in 1930, a group of 1028 economists drafted a statement against the Smoot-Hawley Act that was placed in Congress. It is remarkable how clear-headed this statement was and how it predicted exactly what would happen. Here’s an excerpt:
As Santayana said: “those who cannot remember the past are condemned to repeat it.” Addendum— Pranay KotasthaneI think the Trump administration is indeed reading history—it's just that they are trying to apply an old idea to a vastly different context. I say that because Trump has intentionally avoided referencing the Smoot-Hawley Tariff Acts to justify his love of tariffs. Instead, he refers to the previous instance of tariff raises by the US in the last decade of the nineteenth century. Trump has repeatedly invoked the “tariff king”, President William McKinley, to justify his actions. McKinley was earlier a Congressman who proposed high tariffs in 1890. He championed high tariffs to achieve three goals. The primary goal, as is the case today, was to spur domestic production by reducing imports. The second goal was to protect domestic law and order. As a Congressman, McKinley said, “Reduce the tariff, and labor is the first to suffer”. The theory of change was that more domestic production would create more jobs, reducing anarchy and instability, which were real social problems of those times. The third goal was a reduction in government revenues! Unlike Trump, who believes that the revenue raised through tariffs would be beneficial, McKinley’s government was sitting on surplus cash. This was a problem as excess government expenditure back then was seen negatively (the welfare state idea hadn’t been internalised back then). By increasing tariff rates substantially, he correctly predicted that the revenues collected through tariffs would come down due to a drop in imports. Since tariffs constituted more than 50 per cent of government revenue, a decline in tariff revenue soon ended the government surplus “problem”. The impact of McKinley’s tariffs is less clear than that of the Smoot-Hawley Tariffs. Even the gold standard book on this topic, Clashing Over Commerce: A History of US Trade Policy by Douglas Irwin, states that the tariffs didn’t have a significant negative or positive impact on American manufacturing. At any rate, it is difficult to call it an outright political failure. In a recent Foreign Affairs article, former trade official Aroop Mukharji makes a nuanced point:
Of course, the past is always an incomplete guide for the present. The big mistake in applying the McKinley tariffs or the Smoot-Hawley Tariffs in 2025 is that the world has undergone profound, irreversible changes. Economically, national industries have transformed into long and convoluted global supply chains. Applying the McKinley tariffs today will cause far more damage to American domestic production, as they rely much more on intermediate imports for competitive production. Politically, high tariffs were an accepted component of every nation-state’s toolkit a century ago. However, after seven decades of free trade advocacy, America’s reversion to high tariffs will be seen as a significant betrayal in many countries. Most importantly, US authority has declined—Trump himself believes so. In the 1900s, America was already the world’s largest manufacturing superpower. But today, America’s trade partners have more agency in bilateral trade and the wherewithal to sustain a manufacturing supply chain minus America. It’s for these reasons that Trump’s tariffs are unlikely to work—he’s using a telegram to communicate in the Information Age. P.S.: We must entertain the possibility that tariffs might still lead to some short-term political gains for the Republicans, just like the effect freebies have in the Indian context. Global Policy Watch #2: Going MADInsights on global issues relevant to India— Pranay KotasthaneAn interesting and important AI governance paper was released last week by an influential team that includes Eric Schmidt, a former CEO of Google. Given the pedigree of this team, the paper might have an outsized impact on the future of AI geopolitics and hence deserves a closer look. The paper has attracted attention in media for its opposition to a US-government-run Manhattan Project for building superintelligence. However, that’s not the main import for us living in countries outside the US and China. Titled Superintelligence Strategy, the paper highlights the importance of AI for national security. Its bold claim is that we should treat AI like other dual-use technologies such as “catastrophic dual-use nuclear, chemical, and biological technologies”. It proceeds on the assumption that superintelligence is imminent and hence all methods to restrict its use are fair game. In this regard, it debuts a concept Mutual Assured AI Malfunction (MAIM) on the lines of the Mutually Assured Destruction (MAD) from the Cold War era. This is how the authors describe it:
They argue that MAIM will be the default regime. Nation-states will seek to block the development of an ‘AI-enabled strategic monopoly’ through espionage and Stuxnet-like cyberattacks, kinetic attacks on datacentres(!), and even broader hostilities involving non-AI assets. Given this assumption, they develop a few recommendations. The first one is to not attempt a government-run Manhattan Project which might cause rivals to use the above methods. The second one is a clear agreement between rivals on the escalation ladder of espionage, sabotage, cyberattacks, and kinetic strikes, to ensure deterrence stability. Then, drawing from non-proliferation regimes for biological, nuclear, and chemical weapons, they move on to other recommendations to stop the diffusion of AI. There’s the standard argument for more export controls, on-compute thresholds to disable AI chips remotely, and geolocation and geofencing to prevent unacceptable use of chips. In a striking statement, they write: “the primary goal of compute security is to treat advanced AI chips like we treat enriched uranium.” Finally, they are philosophically against greater open-sourcing of AI models. They want an international agreement to prohibit the release of models exceeding some thresholds, just as is the case with expert-level virologists. Now, as someone sitting here in India, I have several problems with this approach. For one, analogising AI to nuclear weapons is wrong. The former is a general-purpose technology developed by private-sector companies that has the potential to help individuals in all countries in different ways. The other is a specific-purpose technology controlled or financed by governments and requiring expertise for operating it that only trained individuals have. Second, MAD came into being after the world saw the devastation delivered by American nukes on a civilian population, while there’s no such use of AI yet beyond doomsday fiction. MAD was concerned with the use of nuclear weapons and not their development, whereas the authors here are advocating for restrictions on the development of state of the art models. And third, the tonedeafness of the paper is worth noting. Forget US-China competition. The authors are boldly arguing for a strategy that intentionally kneecaps all countries’ AI development in order to preserve American advantage. Now, how do they think the other countries will perceive it? If superintelligence is as imminent as the authors make it out to be, nation-states will find their way around chip export controls through smuggling, hack together low-end chips in large numbers, or simply get access to open-weight models from other countries that have fewer safeguards. If it could happen in a controlled domain like nukes (remember AQ Khan?), this will be pretty much the default in a dispersed technology like AI. To me, this paper advocates a broader shift in American thinking, one that privileges denial over collaboration in AI, and approaches the world with fear rather than confidence. Thankfully, it is not government policy yet, but this paper will move the Overton Window in its direction. Policymakers and analysts in India must watch closely. HomeWorkReading and listening recommendations on public policy matters
|
