Robots and the rise of European superstar manufacturers

Jens Südekum, Joel Stiebale, Nicole Woessner 30 July 2020

a

A

In the US, market concentration has increased in more than 75% of all industries during the last 20 years (Grullon et al. 2019), while average markups have risen mainly because highly profitable firms were able to grasp additional market shares (De Loecker et al. 2020). This elevated market power is, in turn, tightly linked to the falling aggregate labour share of income (Autor et al. 2020). These trends are particularly strong in the US, but they have also been uncovered, though somewhat muted, in other countries (Diez et al. 2019). For example, Andrews et al. (2016) find that global frontier firms – i.e. the top 5% most productive firms within an industry and year - have significantly gained market share relative to laggards across all OECD members. 

An important and yet unsettled question is: what are the underlying drivers of those patterns?1 Explanations for the observed increase in productivity dispersion and the concentration of market power include limited antitrust enforcement and increasing regulation (Gutierrez and Philippon 2018), or increased import competition as a result of globalisation (Autor et al. 2020). But one key explanation, emphasised by The Economist (2016) and many others, is the role of technology.  If newly emerging technological possibilities become available, and accrue primarily to the most productive firms within an industry, those ‘superstar firms’ get even more productive, gain additional market shares, charge higher markups, and earn higher profits. Empirical evidence on technology being a driver for the emergence of this superstar pattern is still rather limited, however. 

In a recent paper (Stiebale et al. 2020), we examine the role of industrial robots in shaping the distribution of firm-level productivity, markups, sales and profits within European manufacturing industries. The global robot market is growing strongly: in 2017, robot sales increased by 21% to a new peak of $16 billion, not even taking into account the cost of software, peripherals, and systems engineering. Robots have revolutionised manufacturing production in many ways, and have become a symbol for novel labour-saving technologies.2 Previous research was mostly concerned with their labour market impacts (Acemoglu and Restrepo 2020, Dauth et al. 2017). We shift attention to productivity, markups, and profits, thereby complementing recent research by Acemoglu et al. (2020) and Koch et al. (2019) on robot adoption at the firm level. 

Empirical strategy

Our empirical analysis combines data on the industry-level stock of industrial robots with firms' balance sheet data for six European countries – France, Germany, Italy, Spain, Finland, and Sweden – from 2004 to 2013. We proceed in two steps. In the first step, we employ recent techniques for production function estimation by Ackerberg et al. (2015) and De Loecker and Warzynski (2012) to measure productivity and markups at the firm level. In the second step, we use those variables to evaluate the effects of robots on the distribution of firm-level productivity and markups within particular industries, countries, and years.

The key result is that the better growth performance (in terms of productivity and markups) of productive and profitable firms is a feature of some, though not all, European manufacturing branches. Investigating which industries tend to exhibit this ‘superstar pattern’, we find that it is considerably stronger in more robotised environments. 

Figure 1 illustrates this finding with an example. The two panels illustrate the evolution of the productivity (TFP) distribution for two different manufacturing industries with vastly different degrees of robotization. The automotive industry (top panel) experienced a spectacular rise in the robot density between 2004 and 2013. For this highly robotised industry, we find that the 90th percentile of the TFP distribution has increased disproportionately compared to the 75th percentile and the median. That is, the most productive automotive producers have increased their productivity much faster over time than the less productive ones. In the bottom panel we show the same productivity evolution for the manufacturing branch of non-metallic mineral products (such as ceramic and glass), in which there was virtually no change in robotisation during the time period 2004 to 2013. In the bottom panel, we find that the upper deciles have actually grown by less than the median. In other words, the ceramic and glass producers which started out with the highest initial productivity levels in 2004 have subsequently seen weaker productivity growth than their less productive counterparts up until 2013. There is no superstar pattern in this (non-robotised) industry. 

Figure 1 Evolution of the firm-level productivity distribution in two manufacturing industries

(a) Automotive industry                   

                    

(b) Non-metallic mineral products

Note: Figure 1 displays the evolution of different percentiles of the TFP distribution from 2004 to 2013, exemplarily for two manufacturing industries which are characterized by a different degree of robotization: motor vehicles (high increase in the robot density, panel a) and other non-metallic mineral products (low increase in the robot density, panel b).

Figure 2 repeats this exercise for the distribution of markups. It again shows the evolution of markup percentiles for two different manufacturing branches. In the automotive industry, where the number of robots per thousand workers has increased substantially, the 75th and the 90th percentiles of markups have grown considerably stronger than the median (top panel). In contrast, this pattern is not present for the electronics industry (bottom panel), which is not strongly robotised. Here, the 90th percentile of firm-level markups has actually strongly decreased in the first years of the sample, and exhibited lower growth than markups at the 75th percentile and the median. 

Figure 2 Evolution of the firm-level markup distribution in two manufacturing industries

(a) Automotive industry                           

            

(b) Electronics

Note. Figure 2 displays the evolution of different percentiles of the markup distribution from 2004 to 2013, exemplarily for two manufacturing industries: automotive industry (high increase in robot density, panel a) and electronics (low increase in robot density, panel b). The percentiles are calculated using sales weighted firm-level markups. Sources: Amadeus, IFR, OECD Stan, Eurostat SBS, own calculations.

Those examples suggest that the emergence of the superstar pattern – i.e. stronger productivity and markup growth in firms that were already highly productive and profitable to begin with – coincides with a higher degree of robotisation. In our paper, we conduct a battery of robustness checks to show that this pattern holds more broadly. Our results indicate that an increase in the stock of industrial robots disproportionately benefits the firms at the top of the productivity distribution. More specifically, robots seem to spur a rise in TFP for the top 20% of firms with the highest initial productivity, but an insignificant effect on the other firms in an industry. The impact on markups also displays considerable heterogeneity. While robotisation negatively affects the markups of firms in the middle and lower tail of the industry-wide distributions, it allows the top 10% of firms to increase their markups even further. 

Digging deeper into the underlying mechanisms, we find support for the theory of endogenous technology adoption. A firm will invest in a productivity-enhancing technology, such as industrial robots, when the expected gains from a reduction in marginal costs are greater than the fixed costs of adoption. Since large firms with higher output and sales tend to benefit more, they might be more willing to incur the fixed costs of investment. Consistently, we find that successful firms not only expand their productivity and markups, but also see a rise in sales and overall profitability, i.e. additional earnings from the robot adoption seem to outweigh the incurred fixed investment costs. 

Finally, we provide evidence that the increased concentration of industry sales that is spurred by the higher exposure to robots contributes to the falling labour income share. In their influential study, Autor et al. (2020) show that highly productive firms are characterised by low firm-specific shares of labour costs in value-added or sales. If, for whatever reason, those firms gain higher market shares, this intra-industry reallocation tends to depress the industry's aggregate labour share. We add to this literature by explicating one particular driver of this pattern: robots, as an example for technological change, seem to have spurred such a reallocation and thereby decreased the industry-wide labour income share stronger in more robotised manufacturing branches.

Policy implications

An increasing dispersion of productivity and markups across firms has broader implications for society. As highly productive firms typically pay higher wages, it may further push up the wages of top earners in these firms, leading to a widening dispersion in household incomes. Perhaps even more important, it may be capital- and firm-owners who benefit most from the recent technological advances. Dauth et al. (2017) provide suggestive empirical evidence that robotisation increases productivity, but not average wages. Our analysis emphasises a key channel through which industrial robots may affect the aggregate labour share: the reallocation of market shares towards successful firms which tend to pay better in absolute terms, but at the same time are able to keep a larger share of revenue as profits.

These trends call for an economic policy approach that supports productivity growth across the broader spectrum, not just among top firms at the technological frontier, and distributes the rents created by new technologies more equally. At the moment, asset ownership and the entitlement to profit earnings are highly unequally distributed. The effects of new technologies on the functional income distribution (higher profit and lower labour income shares) then also imply higher inequality in the personal income distribution. Useful policy steps to counteract those concerning distributional implications could be measures to foster profit-sharing, employee stock options, or similar arrangements. Those instruments would aim for a wider distribution of asset ownership in the society at large.

References

Acemoglu, D, C Lelarge and P Restrepo (2020), “Competing with Robots: Firm-level Evidence from France”, Technical Report 23285, Boston University.

Acemoglu, D and P Restrepo (2020), “Robots and Jobs: Evidence from US Labor Markets”, Journal of Political Economy 128(6): 2188–2244 (see also the Vox column here).

Ackerberg, D A, K Caves and G Frazer (2015), “Identification Properties of Recent Production Function Estimators”, Econometrica 83(6): 2411–2451.

Andrews, D, C Criscuolo and P N Gal (2016), “The Best versus the Rest: The Global Productivity Slowdown, Divergence across Firms and the Role of Public Policy”, OECD Productivity Working Paper No. 5, November.

Autor, D, D Dorn, L F Katz, C Patterson and J Van Reenen (2020), “The Fall of the Labor Share and the Rise of Superstar Firms”, Quarterly Journal of Economics 135(2): 645–709.

Dauth, W, S Findeisen, J Suedekum and N Woessner (2017), “German robots – The impact of industrial robots on workers”, CEPR Discussion Paper 12306 (see also the Vox column here).

Diez, F, J Fan, and C Villegas-Sanchez (2019), “Global declining competition”, VoxEU.org, 2 August.  

De Loecker, J, J Eeckhout and G Unger (2020), “The Rise of Market Power and the Macroeconomic Implications”, Quarterly Journal of Economics 135(2): 561–644.

De Loecker, J and F Warzynski (2012), “Markups and Firm-Level Export Status”, American Economic Review 102(6): 2437-2471.

Economist (2016), The Rise of the Superstars, Special Report.

Grullon, G, Y Larkin and R Michaely (2019), “Are U.S. Industries Becoming More Concentrated?”, Review of Finance 23(4): 697–743.

Gutierrez, G and T Philippon (2017), “Declining Competition and Investment in the U.S”, NBER Working Paper No. 23583.

Koch, M, I Manuylov and M Smolka (2019), “Robots and Firms”, CESifo Working Paper No. 7608 (see also the Vox column here).

Stiebale, J, J Suedekum and N Woessner (2020), “Robots and the Rise of European Superstar Firms”, CEPR Discussion Paper 15080.

Endnotes

1 See John van Reenen discuss the rise of superstar firms in a Vox Video here

2 See the various columns on VoxEU here.

a

A

Topics:  Competition policy Productivity and Innovation

Tags:  superstar firms, robotisation, automation

Professor of International Economics at the Düsseldorf Institute for Competition Economics, Heinrich-Heine University Düsseldorf; Research Fellow, CEPR

Professor of Empirical Industrial Economics, Düsseldorf Institute for Competition Economics, Heinrich-Heine University Düsseldorf

Doctoral researcher, Düsseldorf Institute for Competition Economics (DICE), University of Düsseldorf

Events

CEPR Policy Research