Macro Roundup

Data on superstar firms from The Bloomberg, “…The top 50 companies by value added $4.5 trillion of stock market capitalization in 2020, taking their combined worth to about 28% of global gross domestic product. Three decades ago the equivalent figure was less than 5%....Their median effective tax rate of 35% in 1990 had dwindled to only 17% last year—while profit margins headed in the opposite direction, soaring from 7% to 18% over the same period. They also devote a smaller portion of their earnings to job-creating investments: In 1990, IBM—at the time the world’s biggest publicly listed company—devoted 9% of its revenue to capital expenditures. Fast-forward to 2020, when Apple—its replacement in the top spot—spent just 3%....In 2020 the top 50 companies were sitting on a cash pile of $1.8 trillion, enough to fund their entire capital spending for the year more than five times over….”

Tom Orlik, Justin Jimenez and Cedric Sam, "World-Dominating Superstar Firms Get Bigger, Techier, and More Chinese,"Bloomberg, May 21, 2021, https://www.bloomberg.com/graphics/2021-biggest-global-companies-growth-trends/

World-Dominating Superstar Firms Get Bigger, Techier, and More Chinese

The world’s biggest businesses were doing fine until Covid-19 arrived. Now they’re doing even better.

The top 50 companies by value added $4.5 trillion of stock market capitalization in 2020, taking their combined worth to about 28% of global gross domestic product. Three decades ago the equivalent figure was less than 5%.

In the U.S., President Joe Biden’s administration is seeking to raise corporate taxes as part of a wider effort to halt the long drift to inequality. He wants to reverse at least some of the cuts implemented by his direct predecessor, Donald Trump. He’s also pushing for a global tax deal that would make it harder for the biggest companies to lower their bills by shifting profits to low-tax jurisdictions.

That practice spread as corporations grew bigger. A 2019 study by the International Monetary Fund found that as much as 40% of what on paper looks like foreign direct investment is “phantom investment into corporate shells with no substance and no real links to the local economy.”

In a speech in April, Treasury Secretary Janet Yellen cited a global “30-year race to the bottom on corporate tax rates.” She said agreement among Group of 20 countries on a global minimum charge will create “a more level playing field in the taxation of multinational corporations.”

The U.S. push has drawn resistance from nations such as Ireland, whose low corporate tax rate has encouraged multinationals including Apple Inc. and Google owner Alphabet Inc. to set up regional headquarters there. After initial signs the U.S. might want a 21% minimum rate, the Biden administration has now proposed 15%—a sign of the compromises required to achieve consensus on a contentious issue.

In 1990, there were no Chinese businesses among the top 50 exchange-traded companies; last year there were 8. China’s gains have come largely at the expense of European enterprises, whose presence on the list has shrunk from 15 to 7 over the period.

The extraordinary growth of tech companies in particular is what’s spurring government action. They’re in the crosshairs of politicians and regulators almost everywhere. That includes China, where regulators blocked a proposed initial public offering by Jack Ma’s Ant Group, slapped record fines on affiliates including Alibaba Group Holding, and have extended the crackdown to other tech giants like Tencent Holdings.

In the U.S., there’s bipartisan support for a tougher approach to Big Tech that goes far beyond tax rates. It’s one area where Biden looks set to stick with the policies of his predecessor. The president has nominated Lina Khan, a Columbia Law School professor and author of a landmark paper accusing Amazon of monopolistic behavior, to a key job on the Federal Trade Commission. The FTC is already seeking to break up Facebook Inc. in a lawsuit begun under Trump, and the Department of Justice has filed a monopoly case against Alphabet.

Amazon “has built its dominance through aggressively pursuing growth at the expense of profits,” a strategy that the economics of internet platform markets encourages, Khan wrote in 2017. “Under these conditions predatory pricing becomes highly rational.” For his economic council, Biden has tapped Tim Wu, another Columbia law professor whose 2018 book, The Curse of Bigness, calls for more aggressive use of antitrust law. The growing interest in that agenda has drawn comparisons with the classical age of U.S. trustbusting more than a century ago, when politicians led by Theodore Roosevelt broke up monopolies in oil, railroads, and other industries and subjected the corporate titans of the day to tougher regulation.

Back then, as today, politicians from both parties worried that corporate wealth and power had become concentrated to a degree that was antidemocratic and that a failure to halt the trend might open the way to more radical and populist demands in a society riven by wealth inequalities and a sharp urban-rural divide.

Many of the concerns driving governments are specific to tech and its growing influence in all areas of life, including free speech and the vast amounts of personal data companies accumulate. But others relate to bigness in general, which creates market power: the ability to stifle competitors, strong-arm suppliers, milk customers, and shape regulation.

There’s also a growing body of research showing the increased dominance of superstar firms has placed workers at a disadvantage. Many economists have attributed the slow U.S. wage growth of the pre-pandemic decades at least in part to waning competition. Some tech companies have business models that allow them to scale up without adding many staff. Others, Amazon and Alibaba among them, employ huge numbers of workers but often in low-skill and low-paid jobs—though Amazon, after defeating an attempt to form a union at an Alabama warehouse, has announced pay increases across the board.

Another measure of the growing might of superstar firms is the increased profit margins that Bloomberg Economics documented, which would likely be even wider if some companies weren’t sacrificing short-term income for gains in market share that will deliver larger payoffs in the years ahead.

Economists studying the problem of bigness have concluded it shows up at levels below the world’s top 50, too. For example, a 2018 study found that three-quarters of U.S. industries saw an increase in concentration over the previous two decades, with the market dominated by fewer and bigger companies.

Amid all the concerns triggered by the emergence of superstar firms, the Bloomberg Economics study offers one finding that may be more reassuring. In each of the last three decades, about half of the top 50 spots in the corporate rankings have turned over.

Methodology and Caveats

The Bloomberg Economics study used the EQS function to filter the equity universe for the largest companies by market capitalization for each decade between 1990 and 2020. The screen included all primary securities that were traded as of year-end to account for firms that have since been delisted or acquired. Local currency results were converted into U.S. dollars using historical exchange rates.

To align companies’ financial statements to the calendar year, the study used Bloomberg Query Language (BQL) to capture financial data on a blended annual basis. Where blended annual figures were not available, it used Bloomberg consensus estimates or data for the trailing four quarters. The classification of firms into sectors is based on the Global Industry Classification Standard (GICS) system, with adjustments made to individual companies based on their predominant product or service.

The data provide a snapshot of the start of each decade, which may not capture shifts that occurred in the interim. In addition, Bloomberg Economics examined data for 2019 to assess the impact of Covid on the results for 2020, and found they were broadly consistent with longer-term trends. One calculation where Covid did make a marked difference: the ratio of cash on hand to capex spending jumped higher in 2020. Still, that’s an exacerbation—rather than contradiction—of the longer-term trend.

Ed, just an FYI this is a full writeup on the spillover paper you wanted added to the database. Worth noting their Figure 3 which Ben felt made the spillover effects most clear.

New working paper ~ finds that more research money results in more patents using data from the Department of Energy, “…We providenovel evidence on the magnitude and nature of R&D spillovers by analyzing the US Department of Energy (DoE) branch of the Small Business Innovation Research (SBIR) program. The DoE SBIR program is emblematic of many governments’ venture-like R&D subsidy programs: the funding agency announces priority technologies that they would like to see developed, and then awards competitive grants to the most promising small businesses that apply.2 At the DoE, these priorities range from light sensors, to fuels, to high performance computing software, and are described in Funding Opportunity Announcements (FOAs)….”

Kyle Myers and Lauren Lanahan, "Estimating Spillovers from Publicly Funded R&D: Evidence from the US Department of Energy," Social Science Research Network, April 30, 2021, https://papers.ssrn.com/sol3/papers.cfm

And the spillover are largely domestics, “…Our results indicate that, for every one patent produced by grant recipients due to a marginal increase in SBIR funding, we expect three additional patents to be produced by other inventors across the US and around the world (none of whom directly received a grant). Our results imply that US-based firms and inventors are responsible for roughly 60 percent of the net patents generated by this program.Moreover, many of these new patents are topically quite distant from the technologies initially targeted by FOAs. Taken together, these results suggest that the R&D spillovers from these government grants have a wide impact across both geographic and technological space….Figure 4 also plots the share of patent value captured by all US-based firms and inventors. Our range of estimates spans from about 60% (when all patents are treated equal on average) up to roughly 75%. This suggests that international spillovers might allow foreign countries to capture anywhere from 25-40% of the patent-based value generated by the DoE’s SBIR grants. To our knowledge, this is one of the first direct estimates of the domestic share of returns to public R&D investment…”

They find that R&D spillovers are very large, “…. The firms that receive SBIR funding from the US Department of Energy obtain only 25% of the net patents that their R&D ultimately stimulates. For every one patent they produce, three more patents are produced by others who benefited from their R&D. Even after accounting for marginal differences in patent quality, our estimates suggest that, at most, SBIR firms may appropriate only half of the patent-based value their R&D creates. We provide another piece of evidence that themarginal social returns to R&D may be orders of magnitudes larger than the marginal private returns. Our focus on smaller firms is important since many governments worldwide have R&D policies that target subsidies to small firms. Our focus on public subsidies in the energy sector suggests that this increasingly important sector is fertile ground for the diffusion of new inventions. And while our findings provide another important data point suggesting a large wedge between the private and social returns to R&D, continued work that disentangles externalities and spillovers is essential since it has direct implications for the optimal level of R&D subsidies…”

Bottomline, “..Table 3 summarizes all of these results succinctly. We report the average marginal products and costs and decompose the net patent output of these grants along both the geographic and technological dimensions. We find that grant recipients are responsible for only about one quarter of the net patent output, which indicates geographic spillovers are on the scale of 3× - for every patent they produce, we expect three more from other firms and inventors around the world. Similarly, we find that the nearly two-thirds of net patent output is related to technologies that are not very similar to the original objectives of the DoE’s FOA topics - technological spillovers are on the scale of twofold….”

Table 2 explained, “…Table 2 illustrate the different estimates we obtain when focusing on total funding (federal plus state match), just the match funding, and then just the windfall portion of the match funding.We estimate an elasticity of roughly 2.3 when including total SBIR funding in the model (col. 1), but hypothesize that a significant portion of this variation is driven by the selection bias discussed previously. Focusing just on the state match funding yields an elasticity close to 1.0, which corresponds to 4 patents per $1 million (col. 2). But we are still concerned with the possibility of selection bias here, since match funding, on average, is a direct effect of federal funding. Column 3 reports the results from our preferred specification. Here, we focus just on the windfall funding generated by the match programs, which yields an elasticity estimate of 0.134. This magnitude corresponds to about 0.5 patents per $1 million. Acknowledging the caveats of our focus on the match-based windfall funding, it appears that not accounting for the endogenous allocation of funding overstates the productivity of these grants by an order of magnitude. Columns 4-6 of Table 2 shift the focus geographically outward from grant recipients. We continue to estimate average elasticities between 0.12 and 0.13. But because the base of patent flows grows quickly with each successive grouping (while the elasticity does not decline), we estimate much larger marginal products on a patent-per-dollar basis: 1.4, 1.7, and nearly 3 patents per $1 million for the same-county, US-wide, and worldwide sets of inventors and firms, respectively. These magnitudes suggest that grant recipients might be responsible for only one sixth of the net patent output created by these grants - the R&D spillovers in this setting could be as large as five-fold. In the next section, we relax our assumption of a single bin of investments (by introducing multiple bins b in the set B) to more flexibly estimate spillovers over technology-space….”

Core evidence, “…Figure E.1 contains binned scatter plots of the variation underlying Table 2 in the main text.The axes are scaled so that the linear fits mimic the constant elasticity assumption used throughout the paper. Except for outliers at the funding extremes, this constant elasticity assumption appears very reasonable. The downward shift in the slope of the fitted lines graphically depicts the pattern seen in Table 2 columns 1-3. We estimate smaller elasticities as we shift from using all of the funding variation (federal and state), to focusing on the state match variation, to finally focusing only on the windfall subset of the match-based variation... Table 2 reports the regression estimates of the output elasticities for the four edge cases.Appendix E.1 plots the raw data underlying these estimates, which support our assumption of a constant elasticity.We also report our conversion of these estimates into average marginal products (in terms of patents per $1 million - the approximate size of a Phase II SBIR grant), which are intended to represent the average return on additional investment in an FOA topic. In all cases, we cluster our standard errors at the CPC group level….”

Haltiwanger you wanted added.

New Haltiwanger finds that during recession the churn of the workforce changes, the flow of workers moving from low productivity firms to higher productivity firms downticks and there is an uptick in exits from the labor force from workers employed in low productivity firms.

Core finding, “…We find that in expansions high-productivity firms grow faster primarily by hiring workers away from lower-productivity firms. The rate at which job-to-job flows move workers up the productivity ladder is highly procyclical. Productivity growth slows during recessions when this job ladder collapses. In contrast, flows into nonemployment from low productivity firms disproportionately increase in recessions, which leads to an increase in productivity growth. We thus find evidence of both sullying and cleansing effects of recessions, but the timing of these effects differs. The cleansing effect dominates early in downturns but the sullying effect lingers well into the economic recovery…”

John Haltiwanger, Henry Hyatt, Erika McEntarfer and Matthew Staiger, "Cyclical Worker Flows: Cleansing vs. Sullying," National Bureau Of Economic Research, May 2021, https://www.nber.org/papers/w28802

Impact on productivity, “…Figure 4(a) presents the decomposition of productivity growth into components attributable to poaching and nonemployment flows and shows clear evidence of the cleansing and sullying effects of recessions. On average, worker reallocation through poaching flows contributes 0.1 log points to overall productivity growth each quarter (all statistics on productivity changes are quarterly and have not been annualized). This is a substantial contribution to the overall quarterly average rate of productivity growth of 0.33 when aggregating our micro data.30 However, during recessions there is clear evidence of a sullying effect. In 2006:1 the poaching contribution is 0.13 log points but this declines to 0.02 by 2009:2. In contrast, worker reallocation through nonemployment tends to be a drag on productivity growth, on average, decreasing productivity by 0.67 log points each quarter.31 However, during recessions there is evidence of a cleansing effect since during those times nonemployment flows yield declines in the employment share of low-productivity firms. In 2006:1, the nonemployment component is -0.1 log points but increases to 0.08 in 2009:1. The figure illustrates the staggered nature of these effects in which the cleansing occurs at the outset of the recession-when unemployment rate is rising most rapidly-and the sullying effect peaks relatively further on into the downturn-which the unemployment rate is highest. In addition, the sullying effect lingers well into the recovery…”

Key Evidence, "... The average net employment growth of high-productivity firms is 0.33 percent per quarter with net poaching (the rate at which job-to-job moves reallocate workers to high-productivity firms) averaging 0.27 percent per quarter. In other words, during the 1998-2015 period, job-to-job moves of workers from less-productive employers account for most (80 percent) of the net employment growth of high-productivity firms. The results of the decomposition are also striking for the less productive firms in the industry. In Figure 1(b), low-productivity firms grew at a rate of 0.14 percent per quarter on average from 1998-2015, which is slower than the high-productivity firms. Low-productivity firms lose -0.34 percent employment per quarter from workers “voting with their feet” and moving to firms ranked higher in firm productivity distribution. The positive growth rate for less productive firms is entirely due to strong hiring from nonemployment.In other words, in a typical quarter less productive firms recruit from the pool of unemployed individuals to replace workers moving to better firms. This is also consistent with job ladder models of the labor market. In job ladder models, it is the search and matching frictions that support the presence of low-productivity firms that primarily hire from nonemployment. The patterns of hires and separations in Figure 1 are instructive for understanding the differences in the cyclical dynamics of job-to-job and nonemployment worker flows. Poaching hires and separations both decline for high-productivity firms in contractions with the decline in poaching hires larger so that net poaching declines significantly. Hires from nonemployment decline sharply for low-productivity firms in contractions accompanied by a surge in separations so that net employment growth declines sharply for low-productivity firms. There are similar qualitative patterns for hires from and separations to nonemployment for high-productivity firms but the magnitudes are smaller...."

Bottomline, "... we find evidence of large differences in productivity across firms within the same industry. We also find that more productive firms in the same industry are more likely to grow and less productive firms more likely to contract and exit. The dispersion of productivity across firms is large in magnitude contributing to a high pace of reallocation of workers across firms. Using a decomposition of net job flows into those accounted for by job-to-job flows and those accounted for net flows from nonemployment, we find that much of the overall reallocation of employment from less productive to more productive firms is accounted for by job-to-job flows. The pace at which workers move up the productivity job ladder is highly procyclical. The collapse of the productivity job ladder is consistent with a sullying effect of recessions. In recessions, we find that the reallocation of workers away from less productive firms via nonemployment flows increases. This occurs through a spike in separations to nonemployment along with a decline in hires from nonemployment at low productivity firms. Thus, we also find evidence that this component of reallocation is consistent with a cleansing effect of recessions. The timing of the cleansing and sullying effects differs across stages of the cycle. The cleansing effect peaks relatively early in a downturn coincident with the relatively early spike in separations. The sullying effect peaks later in a downturn but lingers into the early stages of a recovery when unemployment is falling but remains well above trend. Our findings are robust to using a direct measure of productivity based on relative differences in revenue per worker across firms within the same industry and an alternative measure of firm performance based on using the AKM firm premium. Since the AKM firm premium abstracts from worker heterogeneity, this robustness suggests our results are not being driven by variation in the patterns of sorting of heterogeneous workers across heterogeneous firms over the cycle. This is not to suggest that the latter is unimportant but rather that there may be additional effects of cleansing and sullying from sorting above and beyond those we have quantified..”

Lars E.O. Svensson, "Household Debt Overhang Did Hardly Cause a Larger Spending Fall during the Financial Crisis in the UK," National Bureau Of Economic Research, May 2021, https://www.nber.org/papers/w28806

The evidence:“…Thus, the dependent variable is the consumption growth rate from 2006/2007 to 2009/2010, regressed on a continuous measure of debt-to-income (DTI) ratios on 2006/2007. Columns (1), (3), and (5) in table 4.1 report the results (for single and five birth years) in columns (1), (3), and (5) in their table 2. The results establish a negative correlation between pre-crisis indebtedness and consumption growth during the crisis. In isolation, they seem to give some support for the debt-overhang hypothesis. But table 2 of BR does not include any measure of pre-crisis changes in debt among the regressors. Columns (2), (4), and (6) in table 4.1 show that, when the regression is amended to include changes in the DTI ratio from 2003/2004 to 2006/2007, the results are qualitatively similar to the ADJ results in table 3.2. In particular, the debt-overhang hypothesis is rejected, and the regression on only the DTI ratio is subject to omitted-variable bias….”

This is a potentially useful quote, “…A simple model shows that consumption and debt changes are directly and strongly positively correlated, whereas consumption and debt levels are quite weakly negatively correlated….”

Lee Ohanian, Musa Orak and Shihan Shen, "Revisiting Capital-Skill Complementarity, Inequality, and Labor Share," National Bureau Of Economic Research, April 2021, https://www.nber.org/papers/w28747

Takeaway, “…. Our main finding is that the....model continues to fit the data quite well, with high substitutability between unskilled labor and capital, and relative complementarity between skilled labor and capital... The estimation thus chooses a somewhat smaller, though still highly significant, complementarity between skilled labor and equipment capital to attenuate this increase: 0.76 when gross labor share is used, and 0.72 for net labor share, compared to 0.67 in....(gross labor share).... We find that capital-skill complementarity continues to be a critical determinant of U.S. wage inequality, irrespective of which definition of income is used to measure labor’s share. Labor’s share of income does influence the degree of estimated complementarity between equipment and skilled labor, but it does not have a sizeable effect on the model’s ability to account for the skill premium. The findings indicate that capital-skill complementarity remains a quantitatively important determinant of U.S. wage inequality…”

New from Lee Ohanian supports the view that talent is the limiting constraint, finds the rise in inequality btw skilled and lower-skilled workers has been driven by synergies between capital and high skilled workers and substitutability between capital and the lower skilled workers

Core results, "... We find that capital-skill complementarity continues to be a critical determinant of U.S. wage inequality, irrespective of which definition of income is used to measure labor’s share. Labor’s share of income does influence the degree of estimated complementarity between equipment and skilled labor, but it does not have a sizeable effect on the model’s ability to account for the skill premium. The findings indicate that capital-skill complementarity remains a quantitatively important determinant of U.S. wage inequality...."
They look at labors share of gross income and net of depreciation, "...Regarding labor share, we analyze the modelusing both labor share of gross income, which is the measure that has declined, and labor share of income net of depreciation, which has not declined as much, and which is now being analyzed widely within the literature. only considered gross labor share since there were no large changes in depreciation during the years they studied, and thus there was no need to consider both…”

Ufuk Akcigit and William R. Kerr, "Growth through Heterogeneous Innovations," September 20, 2016, National Bureau Of Economic Research, https://static1.squarespace.com/static/57fa873e8419c230ca01eb5f/t/57fd92f590dc5d39708aea59/1476236023160/AK_rev2.pdf

Second, “…incorporate patents and patent citations into our endogenous growth framework, which allows us to connect endogenous growth theory to the empirical innovation literature...Using findings from this empirical literature, we characterize how patent citations would look in our economy and show how citation patterns hold information relevant to the model. While these additions do not impact the model’s economy directly (e.g., firms don’t block rivals with patents), citations provide greater depth to the results that we can characterize. For example, we derive tests that employ patent citations to compare the growth spillover effects from external and internal innovations. Moreover, distributions of patent citations contain much of the information that we need to quantify the model…”

They make three key contributions, “…build a growth model that incorporates multiple forms of innovation, a direct connection from firm size to choices over types of innovations, and multiple step sizes in the impact of innovations that are endogenously determined. Our baseline model analyzes a setting where internal innovations scale up with firm size, while external innovations do not. The tractable model yields analytical solutions and stark predictions about how the innovations of new entrants and small firms will differ from large firms. The model provides micro-founded explanations for small firms experiencing faster average growth and contributing disproportionately to major innovations….”

Bottom line, “…Our model estimates that 26% of aggregate productivity growth is driven by new entry. Of the three-quarters of productivity growth that comes from the action of incumbent firms, the majority of it depends upon external innovation efforts of firms…”

Akcigit/Kerr build a model which looks at innovation drawing an important distinction btw two types of innovation that firms undertake: external and internal, “…Firms undertake external innovations to create new products and capture markets from others, while internal innovations improve product lines that firms currently own….”