Are superstar firms burning too bright?
In January, the pandemic forced the cancellation of TSE’s Digital Conference. Fortunately, the two keynote speakers were able to deliver their lectures remotely. Drawing on his research on innovation and firm performance, John Van Reenen (London School of Economics) examined the rise of superstar firms. Combining economics and computer science, Giacomo Calzolari (European University Institute) focused on the recommendation technologies exploited by many of the 21st century’s new behemoths.
On January 3, Forbes announced that Apple has become the first company worth $3 trillion – greater than the GDP of the UK. The market valuations of tech giants such as Google, Apple, Facebook, Amazon, and Microsoft (GAFAM) have been supercharged by Covid-19, notes John, but their growth began long before the pandemic.
Big firms have been getting bigger, and not just in the digital sector. In 1987, about 28% of US jobs were in firms with more than 5,000 workers; by 2019, it was 35%. Many of the largest firms, like Google, operate without a large workforce so John uses sales as an alternative measure. Across different sectors, he finds rising sales concentration in the US and Europe.
Other dimensions of firm inequality have also increased. “In the UK, for example,” says John, “increases in productivity over the past two decades are happening mainly at the top of the distribution, with the best pulling strongly away from those in the middle. In the US, apart from for CEOs, there has not been much increase in individual earnings inequality within firms. The changes have been between high and low-paying firms.”
John also points to the rise in price markups over marginal cost. “Markups in publicly listed US firms seem to have increased substantially since the early 1980s.. In many countries, we see a reallocation of activity toward superstar firms which tend to have very high markups.”
What explains superstar growth?
For John, today’s superstars were mainly born of technological changes since the 1980s, with some contribution from globalization and institutional change. The GAFAM story reflects the importance of platform competition and network effects in the digital sector. But similar trends are impacting other markets. “The rise of superstar firms may represent the increased importance of fixed costs. Larger firms are better at exploiting intangible capital, such as ICT and software. Wal-Mart has invested billions in software systems to manage just-in-time inventory. There is no way a small independent chain can do that, let alone a Mom ’n’ Pop store.”
Have monster firms emerged because competition is falling, particularly due to the weakening of US anti-trust enforcement? John believes that country-specific institutional factors are unlikely to be the dominant explanation, observing that in some sectors competition may have toughened as globalization, lower communication costs, and trade liberalization allocate greater market share to more efficient firms.
More AI, more market power?
Recommender systems (RS) provide personalized suggestions to users and consumers about specific items or products. They are a collaborative tool designed to predict users’ preferences, using assessments of other users and items. RS are already having a huge impact on markets: for example, 75% of movies watched on Netflix are recommended.
Giacomo’s experimental research uses realistic simulations, operating actual RS in environments in which he can generate preferences and products, as well as controlling the algorithm’s training data. “We find that RS may help consumers, especially when vertical differentiation between products is important. RS induce substantial market concentration, with fewer items achieving positive market share. RS also help to identify superior products when they exist.”
Are superstars good or bad?
For John, it depends on the underlying forces. “Superstar firms are more productive, so reallocation towards them implies higher aggregate productivity. Industries with stronger superstar growth see larger innovation and productivity. Superstars also appear to produce positive productivity spillovers, partly through technology transfer. We found that trading with superstars can increase a small firm’s productivity by 8-10%.”
However, excessive market power can have negative impacts on prices and real wages; productivity and innovation; labor share and social justice. “Since most people’s income comes from their labor, a falling labor share has a direct effect on overall income inequality,” says John. “By using patents and IP to block diffusion, tax arbitrage, and lobbying regulators, are superstars creating barriers to smaller rivals?”
Similar concerns about competition and even democracy have fueled a heated policy debate about RS says Giacomo. “As popular items get a lot of exposure while less popular ones are under-represented, RS risk creating a rich-get-richer affect. As AI algorithms contribute to the generation of the data they are trained on, there are also worries of a cumulative feedback loop.” Compared with exogenous data, his preliminary research finds endogenous data slightly increases concentration but this is only a second-order effect. Most of the biases are instead embedded into the algorithm itself.
Although Giacomo finds biases in the design of RS algorithms, homogenizing products and consumer tastes, his message is positive. The tendency for RS to overstate some items’ quality can reduce competition but the overall effect of RS is to increase competition. “Furthermore, these systematic biases, negatively affecting consumers and some firms, are not due to unavoidable problems, such as endogenous data, and could be eliminated with better algorithms.”
What are the policy implications?
John notes that labor share has fallen less dramatically in the UK than elsewhere. This may be due to a fall in monopsony power driven by an increasingly tough minimum wage. He cites evidence that such policies increase the lowest wages without reducing jobs; although they can squeeze profits, especially for firms with market power.
Rather than imposing costly knee-jerk restraints, or breaking up superstar firms, John calls for modernization of anti-trust policy. “Ex ante regulation can focus on interoperability, data portability and access. Policymakers must consider future competition and market structures when assessing mergers and anti-trust enforcement. The burden of proof needs to shift to the dominant platforms when they acquire smaller firms. And we must find ways to increase structural competition, such as the EU Single Market for Services.”
Technology may be the dominant factor, says John, but labor market policy and institutions – such as minimum wages, collective bargaining, and labor standards for the gig economy – can counterbalance superstar power. Such policy can strengthen job mobility and increase human capital, especially through education and training.
Pictures: Pexels