Macro Roundup

Freddie deBoer, "Is the Conventional Wisdom on Educational Spending All Wrong?," Freddie deBoer, August 23, 2021, https://freddiedeboer.substack.com/p/is-the-conventional-wisdom-on-educational

Is the Conventional Wisdom on Educational Spending All Wrong?

I doubt that anyone would object to the claim that the United States has been trying to spend its way out of educational inequality for decades. Though you sometimes hear, absurdly, that we are “defunding” education, we are spending as much as we ever have, in inflation-adjusted dollars, and that is significantly more than we spent for the great bulk of the history of public education. And this funding increase has been steered in dominant majorities towards “need,” which means poorer schools and disproportionately higher-minority schools2. It’s typical for left-leaning people to lament that poor districts are broadly underfunded relative to others, but it’s hard to justify this belief. Part of the perception problem is that people are operating in a local-spending dominant mindset, when that reality doesn’t really exist anymore.

I really need to underline this point: lower educational expenditures per student can’t be the source of race and income gaps because Blacker and poorer schools receive more per-pupil funding than whiter and richer schools. Sosina and Weathers 2019: “On average, both Black and Latinx total per pupil expenditures exceed White total per pupil expenditures by $229.53 and $126.15, respectively.” Check the tables in the study for more. And this does not encompass the small but growing amount of private dollars that are finding their way into public schools through various grant programs and foundation spending, which likely almost all goes to struggling poor and high-minority schools. This finding flies so directly in the face of the progressive conversation that I find people just can’t hear it, but it actually makes perfect sense. Of course those schools have more funding; we’ve been throwing money at our achievement gaps for 40 years.

(I also urge you to investigate special education funding if you are interested in these topics, as special ed tends to be hideously expensive and devoted to children who have unique difficulties in meeting standards.)

Combine this with the widespread sense that these costs haven’t resulted in better schools and you get some understandable bipartisan exasperation. Frustration about our “failing” schools is potentially misleading; we are arguably one of the best school systems in the world when you match for relevant demographic data in PISA scores, and our median and top 5% of students are competitive with anywhere else in the world. We suffer from truly terrible performance in our bottom quintile which drags our overall numbers way down. But that quintile exists, and it’s the case that we have large and troubling gaps and haven’t closed them despite it being a policy obsession since before I was born. We’ve burned money on it and gotten nowhere. It’s not hard to understand why people conclude that money doesn’t seem to matter much.

This point has frequently been argued through scatterplots, not unreasonably given how consistently they show little or no relationship.

My intent is not to critique either paper, but rather to offer what to me seems like some serious problems we might have interpreting them from a policy standpoint. For now I will take their findings as given: despite the historical failure of spending or spending increases to result in proportional improvements in educational metrics, there is a causal relationship between higher spending and greater performance.

Which is not necessarily a contradiction. Causation does not erase the influence of correlates. For example, all manner of causal variables related to disease are not correlated with infection rates compared across countries because of different levels of sophistication of medical science and other variables. (Hepatitis C causes liver cancer, but not among those who have access to the pricy new medications, which means that as those medicines become more widely available the correlation between Hep C and liver cancer will be reduced.) But this leaves Evans et al. with a pretty glaring question staring them in the face: if school spending leads to student learning… why isn’t school spending leading to student learning? As far as I’m aware not even these researchers dispute the headline national numbers, which show ever-increasing resources spent on schools and nothing to show for it in comparative educational measures. The obvious assumption is that there must be some unexamined variable that is acting as a headwind on the power of additional resources to improve outcomes. But… what?

To my frustration…

1. Neither Evans nor either paper offers a plausible mechanism that could explain the failure of large-scale funding differences to produce the educational gains their research suggests we should observe. If spending more does produce learning gains, why does Vietnam perform as well as the United States? Why does Colorado get equal performance for a small fraction of New York’s costs?

2. Neither Evans nor either paper estimates a threshold of spending past which point student achievement should actually be expected to rise in contradiction with the historical trend, or alternatively, where returns start to diminish.

3. Neither Evans nor either paper can explain what aspects of higher spending actually provide efficient returns, or what this causal relationship actually looks like in simple real-world terms, although Jackson offers useful supposition.

Evans and Jackson both stress that how you spend the money matters. For that matter, Hanushek himself has conceded that targeted strategic spending could move the needle. The problem is that, despite decades of data from armies of researchers, no one really knows what kind of spending actually is strategic3, and the question is so inherently political and contested that it’s hard to imagine a scenario where some sort of useful best practices emerges, especially given that you can make ed data say what you want it to if you try hard enough. I don’t know anyone who says that money never matters or that there aren’t scenarios in which more spending would help. The problem is that we’re talking about big ungainly bureaucracies (yes, even in charter contexts) that have a vast number of gatekeepers and influences, making efficient spending difficult, especially considering that different students have different needs, and we have a profound dearth of compelling evidence for how to spend even if we could do so nimbly. Remember, small effect sizes, huge variances, no or questionable randomization….

Personally, I’m affronted by the fact that there are American schools that, thanks to faddishness and the influence of profiteering, have state-of-the-art computer classrooms but rotting walls full of asbestos. (This school district spent more than a billion dollars on iPads.) To repeat a mantra, the first priority is to make sure that schools are warm, safe, nourishing spaces for children, before we start to worry about quantifying learning at all. But can I prove that my intuitive priorities align with what would raise test scores? I cannot.

As usual, further study is needed. I don’t think the current scholarship is sufficient to outweigh the consistent observation, across many contexts and framed in many different ways, that those that spend more don’t consistently get better quantitative educational metrics in return. I do think that it’s an intriguing line of research. In the short term, though, I can’t see how it impacts the basic political calculus. If I want to spend more on schools, how do I take the current research to the public and ask them to accept tax increases? The political ask remains as hard to make as ever, by my lights. In 1950 we spent less than 3% of GDP on education in this country. Now it’s more like 7%. It’s hard to say that we have better educational or economic outcomes commensurate with that growing price tag. I am a free-spending, let’s-use-the-printing-press kind of a guy. But even I have to ask, how much are we ultimately going to have to spend, and at what point does just cutting people checks become a more efficient option?

1 Hanushek is the husband of the director of Stanford’s much-ballyhooed CREDO project, which…. I’m going to stop before I get conspiratorial.

2 It is worth saying that the view of a struggling American public school student as an urban Black child is not really justified. Because of America’s white majority, a very large portion of struggling students are white, even though their proportion is smaller than in the country writ large. Until recently a plurality of struggling students were white, although both changing demographics and the declining percentage of white students attending public schools have changed that, I believe.

3 I am on record as saying that the best bang for our buck may lie in small group tutoring.

Grossman/Oberfield literature review finds that we don't have a consensus on why labor share in national income has fallen, or if it will continue to decline, stabilize/steady as she goes, or will reverse,"....If we sum the amounts explained by the various mechanisms proposed in the literature, the total easily comes to three or four times the amount that the labor share actually fell.In this concluding section, we offer two potential explanations for this embarrassment of riches…”

“…First, the literature identifies many proximate causes of the decline in the labor share, but precious few fundamental causes. As a result, many authors present different sides of the same coin. This creates a problem of interpretation inasmuch as the same primitive cause can move many endogenous variables together. Even if the various mechanisms that have been suggested all are active,it becomes difficult to gauge what part of the effect estimated in one study has already been accounted for elsewhere. Take, for example, technological advances that give rise to falling prices of ICT capital. Such advances can directly induce substitution of capital for labor, but if they disproportionately benefit larger firms, they can also lead to rising concentration, greater exercise of product and labor market power, and perhaps a change in returns to scale. Without an independent source of variation in the data, it will be difficult to distinguish which effects on factor shares work through the mechanism of rising markups and increased profits and which through altered production techniques. Meanwhile, the nature of technological progress could reflect directed innovation, which in turn could be a response to reduced costs of offshoring. Improved opportunities for offshoring might result from lower communications costs and might also be responsible for deunionization and loss of worker power. In short, the literature on the declining labor share offers a long list of potential mechanisms, many of which seem plausible. The operation of these mechanisms can be studied in a cross-section of firms and industries, and much can be learned about whether microeconomic responses correspond to predictions of the models. But cross-section evidence may not shed much light on the evolution of the aggregates, for the reasons we outlined in Section 3. Meanwhile, the recent economic history is short and macroeconomic variables move together, so identification in the time series raises insuperable challenges. Second, publication bias may be at play. It is widely (though not universally) accepted that the labor share fell in recent decades and that this fact demands elucidation. Many events happened during the period in question, and while some likely have pushed the labor share downward, others may have buffeted the decline. With the goal of finding an explanation for the decline, economists conducting research on the topic may well have pursued mechanisms that predict a decline in the labor share and abandoned those with predictions of countervailing effects. If researchers selectively pursue ideas that imply a fall in the share and neglect those that predict the opposite, the cumulative weight of the theories and their qualifications is bound to overshoot its target….”

Measuring the Decline in the Labor Share

".... Figure 1 shows three popular measures of the labor share in the United States. The BLS headline series for the labor share of the non-farm business sector is the only official series and the most commonly used. This series measures labor compensation plus an estimate of the labor income of self proprietors as a share of gross value added. It gives a clear impression that after decades of stability, or perhaps a modest decline since the late 1940s, the labor share has fallen sharply since 2000, stabilizing in the last decade. In all, this measure shows a decline of about 8 percentage points from its peak in the 1950s. Some researchers prefer measures of the labor share in the corporate sector, because they avoid the ambiguity of proprietors income that we discuss briefly in Section 2.3 below. The figure shows labor compensation in the non-financial sector as shares of gross value added and net value added (which deducts depreciation from the gross measure). Both series show a decline of 6-7 percentage points from the 1980s to the 2010s. The gross labor share was relatively stable through 2000, but then declined to a level below its historical range. In contrast, as noted by Bridgman (2018) and Rognlie (2015), the net labor share had been rising between 1940 and 1980, so the recently lower level is not without precedent. All three series show a modest increase over the last few years, together with a sharp spike in 2020, although it is, of course not yet clear whether this recent history represents a durable rebound or a repeat of the temporary surge during the Great Depression. Another observation that has been emphasized by Karabarbounis and Neiman (2014) is that, over the last few decades, measured labor shares have been declining steadily around the world, with a fall in the global labor share of roughly 5 percentage points since 1975. However, this conclusion remains controversial (GutiÈrrez and Piton, 2020)...."

Will the Labor Share Continue to Fall? (They think the answer is no)

"...But many economists appear to believe that further automation, robotization, globalization, market concentration, and aging of the population spell ongoing declines for the labor share. Some even fear that the labor share in national income might fall to zero…We are more sanguine. The remarkable stability of the labor share for more than a century despite the fact that labor shares in specific industries regularly undergo large changes, and despite the long list of factors that, as we have seen, have the ability to alter the functional distribution of income suggests to us the existence and operation of powerful, stabilizing forces. The exact nature of these forces has yet to be firmly established, but the literature identifies several candidates. The simplest, of course, would be a unitary elasticity of substitution between capital and labor in an aggregate production function, if such a function can be defined. Then any shift in the supply or demand for labor would generate wage movements that exactly counteract the initial shock...."Stabilization could come from the demand side or the supply side of the factor markets...On the demand side, Acemoglu (2002, 2003) points to directed technical change. When the cost share of a factor in scarce supply begins to rise, private agents and an incentive to invest in innovations that reduce that cost. As long as capital and labor are complements in aggregate production, technical change directed at the factor whose share is abnormally high will tend to offset the short-term effects of any shock. In Acemoglu (2003), the supply of capital is perfectly elastic in the long run. Since labor cannot be accumulated but capital ultimately is not scarce, the perpetually rising labor bill induces technical change directed at labor. But the more general point is that a rising factor share for any input will generate a technological response.30 On the supply side, agents might invest in children or schooling to overcome a short-term shortage of some factor. For example....fertility responds endogenously to the return that altruistic parents reap from bearing children relative to the return on their investments in (physical and human) capital. When parents foresee ample earning power for their children, they choose to produce more of them.....individuals spend more time in school when the skill premium is high. In their setting with complementarity between physical capital and raw labor, capital-augmenting technical change or investment-specific technical change induces a short-term decline in the capital share. Then, if the aggregate production function is characterized by capital-skill complementarity, the return to investments in skill rises, leading new generations to remain in school longer. The accumulation of skills in turn raises the return to capital, thereby off setting the initial shock. In the presence of stabilizing forces such as these on the demand or supply side, shocks that move factor shares in the short run are (at least partially) reversed in the longer run. A sudden burst of automation, or a shift in industry composition due to non-homothetic preferences, or differential productivity growth of different factors, or increased competition from imports generates a short-run response in factor shares. But private agents respond to conserve on use of the expensive factors or to provide more of these factors to take advantage of their temporarily high return. Although the conditions for exact long-run constancy of factors shares are delicate in any particular model, it is easy to believe that equilibrating forces might keep these shares in a relatively narrow range. Identifying and measuring these forces in the data is challenging, however, because they operate slowly and may not be seen in cross-sectional data for the reasons discussed in Section 3….”

Gene Grossman and Ezra Oberfield, "The Elusive Explanation for the Declining Labor Share," National Bureau Of Economic Research, August 2021, https://www.nber.org/papers/w29165

Gross versus Net Income, "...National income can be measured gross or net, the difference being depreciation of the capital stock or the natural environment. To the extent that average depreciation rates trend over time say because the composition of the capital stock shifts from more physical to more intangible capital or because society depletes its natural resource base at a faster ratio dividing wage income by gross income will tell a different story about the evolution of the labor share than dividing it by net income.."

Treatment of Intellectual Property, "...Atkeson (2020) takes a different approach to the same problem. Using data for the corporate sector, he computes payments to capital as the sum of payouts to investors and corporate taxes. This method treats all investments as current expenses, in keeping with the suggestion by Barro (2019). His measures of capital income are immune to the recent BEA changes in accounting practices, because payouts to investors are the same no matter whether IP is regarded as an intermediate or a capital good. When he computes the labor share as the fraction of wages in the combined payments to labor and capital, he find like Koh et al. that the current level is low compared to what it was in the 1970ís and 1980ís, but not outside a longer, historical range..."
Treatment of Self Employment and Entrepreneurial Income, "...However, as Smith et al. (2021) point out, entrepreneurs have considerable flexibility when classifying their income in the light of the prevailing tax code. Recent changes in the code have created incentives to reclassify income from forms that give favorable treatment to labor income to those that give favorable treatment to profits, and for frms with high labor shares to leave the corporate sector entirely by forming partnerships. They estimate that reclassification alone accounts for a 1.6 percentage point drop in the labor share in the corporate sector...."

Treatment of Owner-Occupied Housing, "....Real estate constitutes a significant fraction of the capital stock in many countries. For residential housing that is occupied by its owners, statistical agencies impute rent as a return to this capital and include the imputed rent as part of national income. While this practice may be justified as a way of putting residential real estate on a similar footing with commercial real estate, as Rognlie (2015) emphasizes, the returns to owner-occupied housing accrue mostly to workers, not capitalists. Like self-employment income, the issue raised by imputing a return to owner-occupied housing can be avoided at least for the United States by focusing solely on income shares in the non-financial corporate sector..."

Treatment of Factorless Income, "...Oftentimes, the share of capital is treated as a residual, i.e., as one minus the labor share. But efforts can be made to impute the return to capital, which then may not make up the entirety of the difference between national income and wage payments. Karabarbounis and Neiman (2019) label the residual as factor less income Barkai (2020) imputes capital income, using a measure of the risky real interest rate to estimate the rental cost of capital.7 Since the real interest rate spiked in the 1970ís, Barkai finds the imputed capital share to be unusually high at that time and that it has fallen since. This means, of course, that with the wage share also falling, the residual has been growing. But what would account for an expansion in factorless income and who in society benefits from it? Karabarbounis and Neiman (2019) explore three possibilities. First, the residual might an unmeasured return to capital. Second, it could represent pure proof in the face of increased monopoly power and rising markups. Third, it could simply represent measurement error in the return to owners of capital, considering the difficulties that arise in assessing the appropriate risk premium. They finds that none of these interpretations of the residual is fully satisfactory, but that the possibility of measurement error in returns seems closest to the truth. In sum, there remains considerable controversy about the facts. Some believe that the labor share has fallen precipitously while others see a smaller drop or even that the current levels are not far from their historical values. We have tried to argue that many of these differences recast alternative notions of the labor share, and that various measures are valuable for different purposes. Moreover, the significant volatility in all of the measures makes it di¢ cult to distinguish medium turn fluctuations from long-run level changes, or even ongoing trends. Nonetheless, there exists a fair amount of consensus that the labor share today however measured is well below its level of the 1970ís and 1980ís, at least in the United States. We will henceforth take such a decline as a fact worthy of an economists attention...."

Investment-Specific Technological Change, "...even if investment-speciÖc technical change reduced the labor share more in sectors with greater exposure to the change, the impact on the aggregate labor share depends on whether the subsequent increase in the wage o§sets or ampliÖes the initial impulse. This depends on the aggregate elasticity of substitution between capital and labor. Hubmerís conclusion that investment-speciÖc technical change reduced the labor share requires the assumption that capital and labor are gross substitutes in aggregate production. Hubmer imposes gross substitutability by assuming that equipment and structures combine to form a composite capital good that is separable from labor in the aggregate production function. However, other plausible production functionsó such as one that forms a composite of labor and equipment that is then combined with building and structuresó would allow for the possibility that labor and capital are complements. The implications of regressing labor shares on the price of equipment to infer whether capital and labor are substitutes or complements becomes even less clear when we recognize that there are di§erent types of labor and that equipment might complement some (such as skilled labor) while substituting for others (such as unskilled labor)...."

The Rise of Superstar Firms, "...The adoption of new technologies like automated production methods often imposes substantial fixed costs, as Yeaple (2005) and Bustos (2011) have noted. If so, firms that have a larger scale of output will see greater incentives to adopt. Meanwhile, ICT technologies may be more productive in larger firms, if for example, they enhance the integration and coordination of different units (see Lashkari et al., 2021). The differential adoption of new technologies involving ICT and automated production methods have been seen as one reason for a reallocation of resources toward larger firms and for a rise in concentration in many industries. Autor et al. (2020) refer to this phenomenon as the rise of superstar firms....Autor et al. (2020) take a step further. They show that industries that became more concentrated have seen faster TFP growth and a greater increase in patenting. Faster productivity growth in larger firms suggests that these firms benefited disproportionately from recent innovations, perhaps because the adoption of new technologies entails a substantial fixed cost that only these larger firms were willing to incur. In the model they develop to rationalize their findings, larger firms have a lower labor share because they face less elastic demand and choose a higher markup. Together, these forces imply a falling labor share.13 Using establishment-level data for the U.S. manufacturing sector between 1967 and 2012, Kehrig and Vincent (2021) offer a different take on the evidence. They observe firms that, over this period, the manufacturing labor share rose in the median plant. Meanwhile, resources moved to firms with below-average labor shares. However, compositional forces cannot account for the aggregate decline; instead they find that the labor share in these low-share firms fell as they grew in size. Indeed, when they decomposed the change in the manufacturing labor share into the change in the average share, faster growth of firms with below-average shares, and the covariance between the two, they found that the (negative) covariance term accounts for the entire decline...."
Robot Adoption, "...The idea that robots are good substitutes with low skilled labor does not necessarily imply that robots substitute with labor differently than do other the types of capital that appeared earlier. As Acemoglu and Restrepo (2018) point out, it is possible to have technical changed biased toward labor alongside automation. Alternatively, Krusell et al. (2000) show that, in the presence of capital-skill complementarity, a type of capital can be substitute for low-skilled labor but nonetheless complement an aggregate labor input (holding fixed its skill composition). In such an environment, Grossman et al. (2017) show that the subsequent skill acquisition means that declining capital costs can be compatible with long-run stability of the labor share..."
Bottom Line on the Effects of Automation, "...Autor…are well aware of the pitfalls in going from micro to macro. Indeed, their discussion pays heed to the general equilibrium effects that would be overlooked by a strict focus on direct-displacement effects. Even so, their regressions identify deviations of technical change from Hicks-neutrality, whereas aggregate changes reflect deviations of technical change from Harrod neutrality. A cross-section regression of changes in the industry labor share on industry-level TFP growth implicitly holds wages constant. The findings of Autor….are consistent with a pattern of technological progress that is mostly labor-augmenting and an elasticity of substitution between capital and labor less than one. However, aggregate TFP growth also generates a change in relative factor prices that is absent from the Autor…analysis. When technical progress is mostly labor augmenting and the elasticity of substitution is less than one, the wage movements offset the direct effects on the factor shares. The aggregate shifts in the labor share depend only on the extent of deviations from Harrod-neutral technical change in the simple model that we outlined above. The coefficient of -0.579 that they report in their Table 9 from a regression of the labor share on TFP growth is consistent with they hypothesis that virtually all productivity improvements have been labor-augmenting and very little if any have been capital-augmenting..."

Globalization and the Rise of China, "...Finally, we should mention that Autor et al. (2020) included a China-shock variable in their regression of changes on labor share on industry characteristics. They find no evidence that manufacturing industries that were exposed to larger increases in imports from China experienced greater labor-share declines. Growth in China might also have been responsible for booming prices of traded raw materials in the early 2000ís. Castro Vincenzi and Kleinman (2020) show that the decline in the labor share was concentrated in U.S. industries that rely heavily on intermediate inputs. They regress changes in the industry labor share on changes in input prices, finding a negative and significant relationship. According to their estimates, an industry with an average change in its price index of imported materials experienced a 6.2 percentage point decline in its labor share relative to an industry that saw no change in input prices. They also estimate the elasticity of substitution between material inputs and non-material inputs in the industry production functions and find suggestive evidence of strong complementarities between the two, as would be required for a commodity price boom to depress the labor share. For the most part, studies addressing the impact of trade on income distribution have focused on cross-sectional differences in factor shares. As we stressed earlier, to account for the full impact of growing trade on the labor share one must also incorporate the general-equilibrium response of factor prices...."

Increased Product Market Power, "...Alongside (or, perhaps, instead of) technological change and increased globalization, some researchers indict increased product market power as a potential culprit for the declining labor share. As a matter of definition, national income accrues as payments to factors of production or results in pure profits. If pure profits rise, there is less to share among the primary factors of production, including labor. Pure profits could rise due to less stringent enforcement of antitrust laws or because the evolution of technologies gives greater advantage to large firms and forces exit by their smaller rivals. Unfortunately, there is almost no research that ties aggregate trends in product market power and higher pure profits to more primitive causes..."

A Rising Profit Share?'..Barkai (2020) argues that profits shares have been rising in the United States alongside a fall in the shares in value added of both primary factors. To make this argument, he attempted to measure the evolution of aggregate returns to installed capital. An obvious difficulty that he was forced to confront is that firms own most of their capital rather than leasing it, so it is impossible to distinguish rental income from pure profits in either aggregate or firm-level data. In view of this, Barkai sought to impute the required return to capital using a combination of returns on corporate debt and equity, along with an estimate of the depreciation rate.21 He then estimated capital costs as the product of the required return and the value of the extant capital stock. He concludes that the capital share in value added fell by 22% between the early 1984 and 2014, while the share of pure profits in gross value added rose by 13.5 percentage points. Using (5) and an assumption of constant returns to scale, his estimates imply an increase in average markups from 2% to 19% over the period.22 These conclusions are controversial. Karabarbounis and Neiman (2019) point out that the decline in the real interest rate since the 1980ís was preceded in the late 1970ís by a sharp rate hike, which was not, however, accompanied by a rise in the labor share. Moreover, the proÖt share in the period before the spike in real interest rates was higher than it is today. Barkai and others might easily have underestimated growth in the capital share, which would cause them to exaggerate growth of profits. For example, the imputed returns to capital would fall short of true returns, if measures of growth in the capital stock fail to capture all investments in intangible capital for the reasons we discussed in Section 2.5. Also, estimates of required returns might be too low, if depreciation rates have increased with the shift in the composition of the capital stock from mostly structures and equipment to more intellectual property or if market interest rates understate the appropriate discount rates for assessing the user-cost of capital. As Rognlie (2019, p.241) observes about the evolution of markups implied by movements in imputed profit shares since the 1950s, it is not an inconceivable sequence, but it does defy the structural explanations (e.g., market concentration) usually put forward for thinking about markups, none of which should have induced such sharp reversal..."

Rising Markups?"...An increasingly common procedure to estimate average markups, drawing on Hall (1988) and De Loecker and Warzynski (2012), uses together with observed data on some input share and econometric estimates of the elasticity of output with respect to that input. In a much cited paper, De Loecker et al. (2020, henceforth DEU) use firm-level accounting data for a composite input termed cost of goods sold (COGS), which comprises intermediate inputs and a subset of labor inputs that are deemed to be flexible in the short run. Meanwhile, they take capital as a quasi-fixed input. Using techniques popularized by Olley and Pakes (1996) to account for the endogeneity of flexible inputs in response to productivity shocks, they estimate a time-varying production function with variable returns to scale. Finally, they compute a time series for the weighted-average markup by dividing the estimated output elasticity by the cost share of COGS. They conclude that the average markup rose from 21% in 1980 to 61% in 2016, with almost all of the increase attributable to rising markups in the firms in the upper tail of the distribution of price-cost margins.23 Since the markup should be equal to "V =V for any factor V (i.e., the ratio of the output elasticity to the factor share), they conclude that the estimated rise in markups can explain the fall in the labor share...."

Rising Concentration?"...Several studies have pointed to increased industry concentration as evidence of an increase in market power, which in turn could spell a decline the share of revenues available for distribution to the primary factors of production. Barkai (2020) and Autor et al. (2020) find a positive correlation across industries between the rise in concentration and the decline in the labor share. Barkai, in fact, argues that increased concentration (taken to be exogenous) can account for most of the fall in the labor share. Hartman-Glaser et al. (2016) establish that larger firms which have seen the greatest fall in labor share while growing faster than smaller firms tend to operate in more concentrated markets…”

Bottom Line on Product-Market Power, "...If technological developments such as improvements in ICT and automation disproportionately beneÖt larger Örms, as would seem to be the case from the data, then these developments will also enhance the exercise of market power. The appropriate answer to “was it technology or was it market power?” might well be yes..."

Declining Market Power of Workers, "...In addition to a possible increase in product market power, some researchers point to firms ability to extract more of the rents in their relationships with their employees as a source of the shifting distribution of income. A shift in bargaining power may have occurred due to changes in regulations governing labor market interactions, to the sustained decline in union membership in some countries in the post-war period, or to increased concentration of employment within local labor markets that may have allowed firms to exercise greater monopsony power...."

The Decline in Union Membership, "...Consensus estimates put the union wage premium somewhere between 15% and 25% at its height (Rosenfeld, 2014). Farber et al. (2018) show that it has been fairly stable over the last century, with a modest decline over the last few decades. Farber et al. document a positive correlation between state-level labor shares since 1929 and state union membership rates, after controlling for time and state fixed effects and a variety of other controls. They use the passage of the Wagner Act that legalized union organization and the establishment of the National War Labor Board that promoted unionization in establishments receiving defense contracts during World War II as plausibly exogenous sources of variation in states unionization rates and, using these events as instruments, continue to find that labor shares covary with union density. Stansbury and Summers also show a positive correlation across states and industries between the labor share and their measure of labor rents..."

Increasing Concentration of Labor Demand, "... If firms exercise monopsony power in their relationships with their workers, then increasing concentration of firms in the relevant (local) labor markets could account for greater mark downs of wages relative to marginal revenue productivity and perhaps to a smaller labor share. There are now several theoretical models that capture this idea in one form or another, building on Manning (2013). Berger et al. (2019) develop an oligopsony model in which large firms face upward sloping labor supply curves due to heterogeneous worker preferences over the identity of their employer. The firms compete for labor by posting wages, which in equilibrium are below the competitive levels and especially so for the larger, more productive firms..."

Demographics and Education, "...Some authors attribute a portion of the decline in the labor share to changes in the composition of the labor force. The workforce in many advanced countries has been aging due to increased longevity and reduced fertility. And it has been becoming better skilled due to steady gains in educational attainment...."

Potential Drivers of Labor Shares Fall

C. Kirabo Jackson, Does School Spending Matter? The New Literature on an Old Question," National Bureau Of Economic Research, December 2018, https://www.nber.org/papers/w25368

Literature review of impact of increases in education spending finds increase in education spending do have salience, however he the study fails to reconcile the literature with the lack of large scale gains to major increases in spending, “…. In the top half of Table 1, I summarize all of the national-level studies or multi-state studies. Of the 13 multi-state studies, 12 (92 percent) find a positive and statistically significant relationship between school spending and student outcomes. If each study were independent (which is perhaps an implausible scenario), one would not even expect a single study to be statistically significant and positive by random chance. If one were to be conservative, one could treat each source of variation independently. There are school finance studies, tax limit studies, recession-based studies, house-price based studies, and studies that use the roll-out of Title-I. The single study that was not significant was a school finance reform study among which there are several positive estimated impacts. As such, using all five sources of variation, the evidence points to a causal positive impact of increased school spending on outcomes on average. The fact that the early (and less credibly causal) multi-state studies indicate a positive relationship between school spending and student outcomes suggest that the positive association on average is real. The fact that this is also true using only the more recent credible design-based studies that rely on different samples, sources of variation and time periods, is compelling evidence that there is a positive causal relationship and that money does matter on average…”