The largest technology companies affect so many people that it’s natural to compare them to states. That’s true at a simple numeric level—Mark Zuckerberg has more constituents than Narendra Modi, Jack Dorsey has veto power over Donald Trump’s speeches, and Tim Cook has to navigate the perilous China-Taiwan relationship to keep selling phones.

But the parallels are more literal than that. In the COVID-19 crisis, it was tech companies that provided coordination, logistics, and a safety net to mitigate an unparalleled healthcare and logistics crisis. There have been 40.8m unemployment claims filed in the US since the start of the COVID-19 lockdowns, and many companies that haven’t laid employees off have instituted hiring freezes and wage cuts. Amazon, by contrast, hired 175,000 temporary workers in March and April, and increased their pay by $2/hour through the end of May. Since then, Amazon has extended full-time offers to 125,000 of them. Instacart hired 300,000 in March and April, and plans to hire 200,000 more.

In both cases, the companies are not just responding to a crisis, but adapting to a new normal. Internet infrastructure providers have also adapted; AT&T, for example, has handled an increase in bandwidth larger than their network’s total bandwidth in 2014, without service interruptions. Infrastructure users have accommodated larger increases—Zoom’s meeting participant count rose 20x from the start of the pandemic.

This rapid adaptability stands in stark contrast to the way states have responded.

At the start of the outbreak, the state of California had a stockpile of 21 million N95 masks. Unfortunately, these masks were past their expiration date—they’d been accumulated over ten years, since the H1N1 pandemic. California donated the masks to hospitals, but they were reluctant to use them. Fortunately, there were other sources: Facebook donated 720,000 masks. Salesforce sourced 5 million. Apple gave 30 million and manufactured millions of face shields. Flexport didn’t have masks on hand, but bought them anyway, and gave them to the city.

This doesn’t just extend to logistics and personal protective equipment. Companies were also more proactive at enforcing social distancing. Twitter sent employees home on March 3rd. Facebook did so on March 5th.

It even extends to information. A New York Times profile of changes in Facebook’s corporate governance had a buried lede: “On Jan. 27, at a regularly scheduled Monday morning meeting with top executives at Facebook, Mark Zuckerberg turned the agenda to the coronavirus. For weeks, he told his staff, he had been hearing from global health care experts that the virus had the makings of a pandemic, and now Facebook needed to prepare for a worst-case scenario…” Mark Zuckerberg is a busy guy; he’s trying to fend off Google, Snapchat, the U.S. government, the EU, and the media. And he had time to read up on an obscure public health issue in China months before it was on most people’s radars. Jeff Bezos was a bit later to act, but he, too, took more day-to-day control over his company as the crisis ramped up.

Why did these companies succeed where the U.S. federal and state governments failed?

Metrics, Goals, Competence—and Hormesis

Facebook is a business. It has a mission statement (“give people the power to build community and bring the world closer together,”) which doesn’t mean much, but it also produces quarterly financial statements, and those mean a lot. Facebook can frame its decisions in terms of whether or not they serve its goals. Moreover, Facebook’s stock price gives the company signals about how to trade off between the present and the distant future. There’s plenty they could do right now to increase profits at the cost of hurting the long-term viability of the business (those annoying social games were very lucrative, for example). And, by the same token, there are some costs that they’re willing to pay in order to reduce long-term risks.

The masks were part of the latter consideration. Facebook had masks on hand because, in 2018, California had its worst fire season in recorded history. This caused deaths, evacuations, and a drop in air quality—as well as a mask shortage. So Facebook was fortunate: they had a dry run during a natural disaster that cost dozens of lives, and it prepared them for a disaster a few years later that cost hundreds of thousands.

Facebook is not the only institution to benefit from a hormetic dose of crisis ahead of COVID-19. Many of the countries that most successfully addressed COVID-19 were near China—and they acted quickly because of their experiences with SARS and MERS. South Korea had a pandemic plan; Taiwan had procedures; Singapore was ready, too. This raises the question: why do some countries and companies respond to a crisis by planning for the next one, while others, like the state of California, languish? A paradoxical contributing reason for this is that fighting fires is part of the government of California’s job description, so they could respond through conventional channels. In early 2020, when COVID-19 was spreading but not yet visible in the U.S., California was dutifully fighting the last war by raising its budget for fighting wildfires.

The gap between Facebook’s response at one end and California’s at the other showcases the difference between high-agency institutions and institutions ruled by drift and inertia. Facebook is not supposed to be a government—handwavy claims about its 2.6 billion digital citizens notwithstanding, Facebook is ultimately subordinate to the requests of national governments, and doesn’t exercise any form of sovereignty.

It does, however, have part of the incentive structure of good government: since Facebook is so large, and its roster of advertisers is so lengthy, it’s exposed to broad negative and positive externalities. All else being equal, anything that makes U.S. small businesses grow a bit faster will make them spend more money on Facebook; anything that makes it harder for the average person to access the Internet—or that makes the average person less willing to bother—will hurt them. This doesn’t mean that Facebook’s incentives align perfectly with the greater good (they want to grow the pie, but they’re just as happy to take a bigger slice instead). But it does mean that they can act according to their indirect interests rather than direct ones.

Financial markets ultimately help with this, because Facebook is valued highly relative to its sales. Every dollar of incremental sales for Facebook creates about $9 of market value, about $2.50 of which accrues to Mark Zuckerberg personally. And that value is based on Facebook’s long-term outlook; a high-multiple stock establishes a favorable exchange rate between the distant future and the immediate present, giving companies an incentive to make very long-term decisions even at the cost of short-term harms. Facebook, for example, noticed in 2018 that news stories led to vigorous discussion but fewer emotionally rewarding interactions, so they penalized those stories in their algorithm. This hurt their 2018 ad revenues, but benefited their 2030 active user counts, exactly the sort of tradeoff a controlling shareholder would prefer to make.

It’s instructive to note that one part of the government did react swiftly to the crisis: the Federal Reserve cut rates and injected more capital into markets. This was not exactly a comprehensive solution, but given the Fed’s limited toolset and limited mandate, it was the most they could contribute at the time. Dealing with an epidemic is hard, but dealing with an epidemic and a financial crisis at the same time is much harder.

Why was the Fed effective? In part, because it has specific goals: its mandate is to ensure maximum employment, stable prices, and moderate long-term interest rates. In other words, the Fed can frame every decision it makes in terms of how it achieves specific, measurable goals. An economic collapse brought about by COVID-19 could cause deflation (as revenues decline, businesses would lay off employees and default on debts, leading to a deflationary cascade) or inflation (a drop in production and a run on consumer goods could make prices spiral upward). By acting quickly to stabilize the market, the Fed narrowed the immediate scope of possible outcomes.

Operationally, the Fed functions differently, too. Its deliberations mostly occur in private, with a small group of people. It’s uncomfortably non-transparent, but it does mean they can hash out the argument for or against an extreme policy move, and then present it as a finished product. Legislation develops more publicly; since a bill can’t pass without widespread support, the contents of a new bill will leak as its sponsors try to corral backers—and anyone who dislikes the bill (or some part of it) can produce damaging leaks to scuttle it. The twelve members of the Federal Open Market Committee can convene on a single conference call, but the 218 House members necessary to pass a bill can’t.

The Fed was also effective because they, too, had received a hormetic dose of crisis. While COVID-19 is primarily a health crisis, its secondary effects could have quickly led to a financial crisis. The Fed and other central banks had experience dealing with such crises, and they knew that a hit to the real economy can be compounded by a collapse of the banking system. The mistake they made in 2008 was to view the subprime mortgage crisis as “contained” because it was small; it turned out to produce enough uncertainty to lead to a much larger financial crisis. Rather than risk repeating the mistakes of 2008, the Fed quickly cut rates, and signalled that it was ready to engage in unconventional economic stimulus.

Institutional Autopilot

In October 2019, Johns Hopkins Center, the Nuclear Threat Initiative, and the Economist Intelligence Unit released a pandemic preparedness index, ranking every country in the world in terms of its ability to handle a deadly disease outbreak. The United States ranked first in the world. On paper, the US was prepared; there were government and non-profit organizations whose mandate included pandemic preparedness, and the U.S. spends $3.6 trillion on healthcare. The budget was there. What was missing was agency: since a pandemic is a rare event, people at the CDC did not get promoted on the basis of preventing plagues so much as on the basis of being the sort of person who seemed like they’d be good at plague prevention.

Idealists believe that every institution exists to build a better future for its participants. Cynics think every institution exists to perpetuate its own existence for the current benefit of whoever is in charge. The realist approach is to accept the cynical view as the ultimate asymptote to which all institutions trend, and to create incentives that align cynical self-preservation with prosocial behavior. Amazon employees who worked extra hours and took personal risk to keep delivering packages weren’t doing it to keep the economy afloat; they wanted bonuses, and they didn’t want to get fired. Facebook and Twitter didn’t send employees home out of pure altruism; they did it because operational disruptions are bad for business.

The economics of technology companies encourage them to insert themselves into the highest-margin gap they can find between supply and demand. But this gives them a peculiar form of altruism: what’s good for the country is good for Big Tech, because Big Tech gets a disproportionate share of the upside. And because of how their stocks are valued, the long-term is what matters. Most of the net present value of a growth stock is far in the future, so they’re deeply sensitive to systemic risks. The paradoxical combination of risk-tolerance (they have to continuously reinvent themselves to grow) and risk-aversion (Amazon trades at over 100 times earnings, so its value is almost entirely based on what happens a long time from now) basically forces tech companies to be high-agency.

Institutions can stay on autopilot for a long time. Especially in the U.S., an extraordinarily wealthy country whose government is keenly sensitive to the needs of powerful interest groups, it’s possible for many business and political leaders to shrug off distant systemic risks. But lower demand for agency did not significantly impair the supply of agency. Growing up, Jeff Bezos dreamed of space colonies, and Mark Zuckerberg obsessed over Roman orators. But Bezos didn’t get a job at NASA, and Zuckerberg didn’t run for office; they found a better place to be ambitious. Expertise doesn’t always transfer across domains, but agency does. And in a crisis, the people in charge aren’t determined by org charts and delineated responsibilities, but by who acts first.

This is a relatively recent phenomenon. Had Bezos been born a few decades earlier, his obsession with space travel would have led him to getting a job at NASA, because NASA was one of organizations in the world that could plausibly get humans into space, and to join the other one you had to speak Russian. NASA today has a budget and a staff; NASA in the 1960s had those things, but it also had a plan, a deadline, and a meaningful competitor. The plan to land a man on the moon was arbitrary, and the deadline was a nice round number, not an informed estimate of how long the project would take. It would be hard to have an informed estimate in 1961, when Kennedy promised to put a man on the moon by the end of the decade. Much of the necessary technology hadn’t been invented yet, so estimating production timelines was fraught.

During the COVID-19 epidemic, we’ve all learned to think more deeply about immune systems and infectious disease. It’s worth thinking about the memetic immune system. The U.S. government has few natural vulnerabilities; the widely-known gap between Congress’ 15-25% approval rating and its 98% incumbent reelection rate shows that popular approval is not strictly necessary. But even the largest technology companies are deeply vulnerable to outside threats: everything Apple sells will be obsolete in a few years; Facebook’s core mobile ad product generated $0 revenue when they went public; Amazon has faster-growing competitors in every business it’s in. So, tech companies have a vigorous immune response to threats; they’re constantly encountering them, evaluating, and responding, often writing the rulebook as they go.

The U.S. government doesn’t have such a robust immune system; it can remain oblivious to problems almost indefinitely because almost no problem represents an existential risk. In fact, the internal incentives of the U.S. government actively encourage low-grade problems: any harmful social issue or geopolitical threat represents job security for whoever is responsible for addressing it, and solving problems means losing one’s justification for working. Prior to the COVID-19 epidemic, the single largest increase in unemployment in U.S. history was the result of the government solving a problem: in September 1945, unemployment rose by nearly two million as the US demobilized.

A problem like winning a war is visible and measurable, and it creates incentives that overcome institutional inertia. Smaller problems, though, don’t affect this. The U.S. slowly losing its domestic manufacturing base and international relevance is an issue, but it’s not an issue with a discrete win/loss condition, just one with a long-term trend. Disability fraud, corporate welfare and tax avoidance, expensive healthcare and education—all of these problems fit the same framework. There’s no external threat, they worsen at a smooth cadence instead of in discrete steps, and attempting to solve them without ever succeeding is a sinecure.

The memetic immune system is the set of institutional practices and individual habits that lead an institution to accept some ideas and vigorously reject others. It can only be trained through risk; if a company accepts the idea of managers without metrics, or of employees who can’t achieve goals but don’t suffer consequences, it quickly falls behind. But a rich government is too hard to analyze at that level, and incentives are too diffuse. When institutional sclerosis is everyone’s problem, it’s nobody’s problem.

Technology companies responded fast because they’re in the habit of acting. Other institutions responded slowly, because they’re in the habit of reacting, and a novel infectious disease did not have a default institutional reaction.

Ultimately, there’s a teleological difference between these institutions. Tech companies have a defined purpose, and they’re a vehicle for achieving that purpose. Google wants to organize the world’s information, and turning a gigantic profit while doing so is one of their operating constraints. SpaceX wants to make humans a multi-planetary species, and submitting competitive bids for rocket launches is an early step to accomplish this.

But the U.S. government doesn’t have a defined purpose, just a long list of functions, and those functions aren’t regularly interrogated. It’s uncomfortable for modern Americans to ask what the ultimate goal of the government is. The very idea is both utopian (do we really think the U.S. government can accomplish it?) and non-democratic (can 330 million people agree on a single goal?). But an institution without an ultimate purpose is functionally dead. The only question is how long it takes for entropy to win.