Monday, August 5, 2019

Some Thoughts on Markups

A group of recent research studies have argued that "markups" are on the rise. As one of several prominent examples, a study by Jan De Loecker, Jan Eeckhout, and Gabriel Unger, called "The Rise of Market Power and the Macroeconomic Implications" presents calculations suggesting that the average markup for a US firm rose from 1.21 in 1980 to 1.61 in 2016 (here's a working paper version from Eeckhout's website dated November 22, 2018). The Summer 2019 issue of the Journal of Economic Perspectives discusses the strengths and weaknesses of this evidence in a three-paper symposium:

Here are some of my own takeaways from these articles:

1) For economists, "markups" are not the same as profits. Profits happen when price is above the average cost of production. Markups are defined as a situation where the price is above the marginal cost of production. This definition is also why markups are so hard to measure. It's pretty straightforward to measure average cost: just take the total cost of production and divide by the quantity produced. But measuring marginal cost of production is hard, because it requires separating a firm's expenses into "fixed" and "variable" costs, which is a useful conceptual division for economists but not how firms divide up their costs for accounting purposes.

2) There are a number of economic situations where there can be persistent markups while profits remain at normal levels. As one example, Every intro textbook discusses "monopolistic competition," a situation in which firms in a market sell similar but differentiated products. Examples include clothing stores in the same shopping mall with different styles, gas stations with different locations, products sold with different money-back guarantees, and all the everyday products like dishwasher soap. The basic textbook explanation is that in a setting of monopolistic competition, firms will be able to set prices above marginal cost, charging more to consumers who desire the specific differentiated characteristics of that product. However, part of the definition of monopolistic competition is that other firms can easily enter the market and expand production. The result is a situation of positive markups (price higher than marginal cost) but only average profits.

3) Another example of positive markups arises in companies with high fixed costs and low marginal costs: as a simple example, think of a video game company where the cost of creating the game is high, but once the game is created, the marginal cost of providing the game to an additional user is near zero. It's quite possible for this kind of company to have a positive markup (price over marginal cost), but also to suffer losses (because price isn't enough above the markup to cover the firm's fixed costs).

4) Many big tech companies--Facebook, Amazon, Google, Apple, Microsoft, and others--have a number of products that share this property of high fixed costs and lower marginal costs. Thus, these companies are likely to have high mark-ups. Moreover, given their technology, size, and business, practices, it may be difficult for entrants to challenge these companies. As a result, companies like these may have an ability for a combination of sustained high markups and high profits over time. For example, in the study mentioned above by De Loecker, Eeckhout, and Unger, the overall rise in mark-ups not because of a rise in markups for the median company, but because of a very sharp rise in markups for a much smaller number of companies.

5) An emergence of high-markup, high-profit firms may be in some cases a positive step for productivity and wages. There is an emerging literature on what are sometimes called "superstar" firms (here's a recent working paper version of "The Fall of the Labor Share and the Rise of Superstar Firms," by David Autor,  David Dorn, Lawrence F. Katz, Christina Patterson, and John Van Reenen ). Imagine a company that makes large-scale investments in information technology, both for the logistics of running its operations and for quality control, and also in widespread use of far-flung supply chains. Based on these high fixed costs, such a company may be able to expand in the market, taking market share from smaller firms. But economists commonly hold that when a company succeed by offering the products that consumers desire at attractive prices, this is a good thing. Thus, antitrust authorities need to think carefully about whether companies with high markups are gaining  high profits through pro-consumer fixed investments or by anti-consumer restraints on competitors. 

6)  It's hard to measure whether markups are rising for the economy as a whole. Detailed studies of a certain firm or industry look carefully at the production process for that firm or industry and estimate fixed and marginal costs. But how can such data be collected for most companies in an economy?  Some approaches use firm-level accounting data, or data on capital investment and depreciation across industries, or sectoral-level data on outputs and inputs. Using this data to estimate production functions across firms and markups involves a number of modelling choices. It turns out that assumptions like whether industry is producing with constant returns to scale or increasing returns to scale matter a lot. There are a bunch of genuinely hard and disputed question on how to proceed here. 

7) As an example, one approach is to rely on accounting data. The prominent study by De Loecker, Eeckhout, and Unger uses accounting data (from Compustat) which breaks down costs of production into two main categories: "Cost of Goods Sold" and  "Selling, General, and Administrative." If one thinks of the Costs of Goods Sold as a proxy that captures variable costs, one can then use this data along with a measure of the fixed capital stock of firms to do more detailed calculations, which with some assumptions (like assuming that all profits are paid to owners of capital) will  imply estimates for markups. But there are a bundle of underlying assumptions here. For example, Cost of Goods Sold is defined differently by the accountants across industries, and it seems to be a falling share of total costs over time. Sorting out the underlying economic meaning of the accounting data, along with assumptions are necessary and what implications these assumptions have, is an ongoing area of research. 

8) Other attempts to measure markups often rely on measuring a firms's amount of capital, which could be thought of as a fixed costs, and then looking at the profits received by owners of capital. But many firms have both "tangible" capital like machines that is relatively well-measured, and also "intangible" investments, which include situations where a firm has made past investment in knowledge (like research and development) or organizational capabilities that are now paying off. Of course, measuring intangible investment (and thinking about how fast it might depreciate) is very hard, but there's some evidence that intangible investments have become more important over time. But if firms with high markups and profits are largely benefiting because they made large investments in the past--in the form of intangible capital--then this should probably be viewed as a positive for the economy. 

9) Estimates of markups over time often reveal patterns that raise additional questions. For example, the study by Loecker, Eeckhout, and Unger finds that most of the sharp rise in markups happened among a smaller segment of firms in the 1980s and 1990s, and hasn't changed much since. (How does that timeline fit with one's internal narrative about what has caused higher markups, and in particular a belief that the causes are relatively recent?) Other studies that project back further in time suggest that markups were especially large in the 1960s. (So perhaps we need multiple explanations for what affected markups then and now?) 

10) If very large rises in markups have occurred, then they should have implications for other areas of the economy. For example, Basu points out in his JEP paper that it's conceptually possible to draw a connection from higher corporate markups to labor receiving a lower share of income (a fact discussed here and here). But high estimated of the rise in markups suggest that the labor share of income should have fallen by much more than actually observed. Or as Syverson points out in his JEP paper, a rise in markups  implies either that prices have risen quickly or that (marginal) costs have plunged. Low inflation rates means that prices have not risen quickly, and evidence that costs have plunged economy-wide is scanty. Thus, both authors express a view that while it is it plausible that markups have risen, the size of such a rise must be relatively modest to fit with other observed economic changes. 

There's much more in the articles themselves about methods of measuring markups, what methods produce estimates that are higher or lower, possible connections (or not) to concentration of industry. Attempts to measure whether markups are rising in the US economy, and if so, by how much is a live and active area of research. If someone tells you that a very large rise in markups is a settled fact, they are showing a lack of awareness of the actual evolving state of play in this literature.