Bottom-Up Markup Fluctuations
GAICIL Insights
Picture an antitrust dashboard with one gauge.1 The needle climbs, and the alarm sounds: markups are up, competition must be down. Bottom-Up Markup Fluctuations takes that dashboard apart, wire by wire, and finds that the single gauge is wired to several engines running in different directions. In a granular macroeconomic model with oligopolistic competition, firm, sector and aggregate markups respond to the same firm-level shocks in different ways. The model reproduces those patterns in French administrative firm-level data, from the bottom of the economy to the top.
Start with the firm. When a firm gains market share, its markup rises. When a sector grows more concentrated, the sectoral markup rises too. Then the picture splinters. The relationship between markups and sectoral output varies by firm size: negative for small firms, positive for large ones. Sector-level markups move positively with sectoral output, while sectoral markups show no systematic relationship with aggregate output. Read one by one, these facts look like a contradiction. Read together, they are the fingerprint of very different firms reacting to their own shocks inside concentrated markets.
Now run the thought experiment that matters for artificial intelligence. A large developer ships a markedly better model. Users flock to it. Its share jumps, its markup jumps, and the measured markup of the whole segment jumps with it, all while output expands and customers get a superior product. The gauge reads danger. The mechanism reads success. A regulator watching only the aggregate would see entrenchment where the paper's framework sees a productive shock propagating through an oligopolistic structure.
Call it the success paradox of concentration metrics. The better the winning product, the worse the headline number can look. Frontier AI markets are granular almost by design: a handful of large developers, a ring of specialists, a long tail of small entrants, all in the same segments. In exactly those markets, aggregation compresses heterogeneity into one figure and hides offsetting movements. Small firms in an expanding sector may see their markups fall while large firms see theirs rise. Averages erase the very question that matters: who is winning share, and why.
The paper carries a second surprise for macroeconomists. Its granular setting produces nonnegligible aggregate fluctuations in output and markups. Large firms are big enough that their idiosyncratic shocks refuse to wash out. Translate that into an economy increasingly shaped by a few AI-intensive firms, and the business cycle and the competition file merge into one. A breakthrough, or a stumble, at a single giant can ripple into aggregate output. Competition enforcers and central bankers end up staring at the same firms.
The authors are candid about what their static benchmark leaves out. Markup movements in the model come from changes in the shape of the demand curve after firm-level shocks. The conclusion flags price-setting and customer-accumulation frictions, aggregate monetary and financial shocks, and richer intertemporal markup strategies as prime extensions. It also points to multiproduct firms linked through intermediate inputs, some competing locally and others nationally or internationally. Each of these reads like a description of AI markets: lock-in, data accumulation, and layered supply chains for compute and models that shape pricing over time.
What should an enforcer do with this? Treat the markup as a symptom that needs a diagnosis. A rising markup in a sector led by an expanding, productive firm calls for a different response than a rising markup sustained by barriers to entry. The paper does not decide where any particular AI market sits on that spectrum. It hands over the instruments to ask the question correctly, and it shows why one aggregate series is a poor guide to the answer.
The deeper payoff is a vocabulary for distinguishing two things that look identical on a chart: the markup of success and the markup of entrenchment. One rewards the firm that built the better product. The other rewards the firm that blocked the door. In an economy where innovation increasingly arrives through a few large firms, confusing the two can be a costly measurement error. Granular, firm-by-firm evidence is the cure, and this paper shows how to build it.
The practical agenda follows naturally. Collect firm-level data at the level where shocks actually hit. Track share shifts alongside markups. Ask whether expanding firms are lowering quality-adjusted prices for users. Only then decide whether the needle on the dashboard signals a problem worth fixing.
Introduction
We study markup cyclicality in a granular macroeconomic model with oligopolistic competition. We first characterize how firm, sectoral, and aggregate markups comove with output at different levels of aggregation in response to firm-level shocks. We quantify the model’s ability to reproduce salient features of the cyclical properties of measured markups in French administrative firm-level data from the bottom (firm) level to the aggregate level. We document that (i) firm-level markups rise with market share and sector-level markups with concentration, (ii) the relationship between markups and sectoral output varies by firm size—negative for small firms but positive for large ones, (iii) sector-level markups move positively with sectoral output, and (iv) sectoral markups show no systematic relationship with aggregate output. Our model helps rationalize these seemingly conflicting patterns of markup cyclicality in the data.
Takeaway
We examine markup cyclicality through the lens of a simple oligopolistic macroeconomic model with rich implications for the behavior of markups at the firm, sector, and aggregate levels. Working with administrative firm data for France, we show the model can reproduce qualitatively, and many times quantitatively, an array of markup-related empirical moments at various levels of disaggregation. In our framework and measure of markups, we can reconcile seemingly conflicting variants of “markup cyclicality” that have been considered in the literature. Finally, our granular oligopolistic setting produces nonnegligible aggregate fluctuations in output and markups.1
One obvious route for future work is to superimpose in our model alternative shocks and frictions. Along this line, prime candidates would be to consider price setting and customer accumulation frictions (see Gilchrist et al. 2017; Hong 2017; Afrouzi and Caloi 2024),2,3,4 as well as aggregate monetary and financial shocks. Relatedly, we have focused on static, intratemporal markup decisions in which movements in markups are the result of changes in the shape of the demand curve in response to firm-level shocks. These forces would remain relevant even if one were to extend the model to allow for richer intertemporal dynamics that result in more complex dynamic markup strategies (see Rotemberg and Saloner 1986).5 Bringing the resulting firm, sector, and aggregate dynamics to data—and comparing them against the forces in our static benchmark—would render possible an assessment of the empirical merits of this more general approach.
Finally, extensions to more realistic and richer product and market structures would allow us to more accurately map model objects to the increasingly detailed micro data available to researchers. Such extensions include multiproduct firms, interlinked through intermediate inputs, with some firms competing only locally in spatially segmented (product and factor) markets and others doing so nationwide or internationally.
Notes
- Ariel T. Burstein, Vasco M. Carvalho & Basile Grassi, Bottom-Up Markup Fluctuations, 140 Q.J. Econ. 2619 (2025), https://doi.org/10.1093/qje/qjaf029. Takeaway source: VIII. Conclusion, reproduced from the openly licensed version (CC BY 4.0). ↩
- Simon Gilchrist, Raphael Schoenle, Jae Sim & Egon Zakrajšek, Inflation Dynamics during the Financial Crisis, 107 American Economic Review 785–823 (2017), https://doi.org/10.1257/aer.20150248. ↩
- Sungki Hong, Customer Capital, Markup Cyclicality, and Amplification, Federal Reserve Bank of St. Louis Working Paper No. 2017-33 (2017), https://doi.org/10.20955/wp.2017.033. ↩
- Hassan Afrouzi & Luigi Caloi, Endogenous Firm Competition and the Cyclicality of Markups, 106 Review of Economics and Statistics 1638–1654 (2024), https://doi.org/10.1162/rest_a_01281. ↩
- Julio J. Rotemberg & Garth Saloner, A Supergame-Theoretic Model of Price Wars during Booms, 76 American Economic Review 390–407 (1986). ↩
Cite as: Ariel T. Burstein, Vasco M. Carvalho & Basile Grassi, Bottom-Up Markup Fluctuations, 140 Q.J. Econ. 2619 (2025), https://doi.org/10.1093/qje/qjaf029
Header image: Grocery Store Aisle, vermont.jpg by Tessa Bury, CC BY 4.0, via Wikimedia Commons.
