AI and productivity

I’m moderating an event on digital technology and productivity later this month, and Noah Smith just published a great column on the topic, based largely off of a new paper by Erik Brynjolfsson, Daniel Rock and Chad Syverson. Here’s a key bit:

Often, when a very versatile new technology comes along, it takes a while before businesses figure out how to use it effectively. Electricity, as economist Paul David has documented, is a classic example. Simply adding electric power to factories made them a bit better, but the real gains came when companies figured out that changing the configuration of factories would allow electricity to dramatically speed production.

Machine learning, Brynjolfsson et al. say, will be much the same. Because it’s such a general-purpose technology, companies will eventually find whole new ways of doing business that are built around it. On the production side, they’ll move beyond obvious things like driverless cars, and create new gadgets and services that we can only dream of. And machine learning will also lead to other new technologies, just as computer technology and the internet led to machine learning.

This is the idea behind Michael Hammer’s vision of “reengineering”:

It is time to stop paving the cow paths. Instead of embedding outdated processes in silicon and software, we should obliterate them and start over.

It’s also closely related to Rebecca Henderson’s idea of architectural innovation; she argues that incumbent firms are quite bad at changing their architecture in response to new technologies.

If you combine the power of reengineering with the idea that incumbents struggle to do it, you end up with something like Chris Dixon’s full-stack startup idea, which, sure enough, others are applying to machine learning companies.

All of which is to say, I find Brynjolfsson et al’s theory quite plausible. Machine learning will become more valuable as it is incorporated into how organizations are designed, rather than just inserted into current structures. (It’s also why I think that even if AI progress slows, business will still shift dramatically.)

That’s also why I think Brynjolfsson and Andrew McAfee’s latest book, Machine, Platform, Crowd is so valuable. Redesigning organizations isn’t just about machine learning; when you combine ML with crowdsourcing and other newer models, you end up with fundamentally different kinds of organizations, like Numerai:

Numerai is a hedge fund managed by an anonymous community of data scientists. It encrypts its data and allows anyone in the world to continuously apply machine intelligence to the set and anonymously submit price predictions back. Numerai turns these predictions into trades and compensates the best performing models with bitcoin.

One open question: how does the wild divergence of productivity between firms fit in, given that it’s driven largely by digital technology? Is it the case that the winners so far are the ones who’ve really organized around these technologies? Or are they just better at the lesser early adoption that barely moves the needle, and will be toppled by a new era of AI-full-stack startups?

Notes on redistribution

Some entrepreneurs and some libertarians (or “liberaltarians”) appear to be warming to redistribution and the welfare state. But there’s a reexamination happening in left-of-center policy circles, too. I offer no opinion on that conversation here, but want to clip together a few references…

The limits of redistribution

Franklin Foer on Elizabeth Warren and the future of the Democratic party:

Nor is Warren’s driving obsession wealth redistribution. That’s important politically, because many Americans simply don’t begrudge wealth, and “inequality” as a clarion call hasn’t stuck… Rather, Warren is most focused on the concept of fairness. A course she taught early in her career as a law professor, on contracts, got her thinking about the subject. (Fairness, after all, is a contract’s fundamental purpose.) A raw, moralistic conception of fairness—that people shouldn’t get screwed—would become the basis for her crusading. Although she shares Bernie Sanders’s contempt for Wall Street, she doesn’t share his democratic socialism. “I love markets—I believe in markets!” she told me. What drives her to rage is when bankers conspire with government regulators to subvert markets and rig the game. Over the years, she has claimed that it was a romantic view of capitalism that drew her to the Republican Party—and then the party’s infidelity to market principles drove her from it.

A lengthy piece in Democracy:

Suppose we raised marginal tax rates on the highest income households from 39.6 percent to 50 percent… the increase would raise taxes by an average of $6,464 for those in the 95-99th percentiles (those with average incomes of $321,000 in 2013). Households in the top 1 percent (with average incomes of $1.571 million in 2013) would pay an additional $110,968 and those in the top 0.1 percent an additional $568,617… Now imagine that all of the revenue collected from this change was distributed evenly to the bottom 20 percent. The total revenue raised is $95.6 billion and allows each household at the bottom to have an extra $2,650 in post-tax income.

Although not directly discussing redistribution, another relevant estimate comes from David Autor in Science:

Between 1979 and 2012, the share of all household income accruing to the top percentile of U.S. households rose from 10.0% to 22.5% (89). To get a sense of how much money that is, consider the conceptual experiment of redistributing the gains of the top 1% between 1979 and 2012 to the bottom 99% of households (10). How much would this redistribution raise household incomes of the bottom 99%? The answer is $7107 per household—a substantial gain, equal to 14% of the income of the median U.S. household in 2012. (I focus on the median because it reflects the earnings of the typical worker and thus excludes the earnings of the top 1%.)

He goes on to say that, in terms of total dollars, the rise in the college wage premium has been more significant:

This increase in the earnings gap between the typical college-educated and high school–educated household earnings levels is four times as large as the redistribution that has notionally occurred from the bottom 99% to the top 1% of households. What this simple calculation suggests is that the growth of skill differentials among the “other 99 percent” is arguably even more consequential than the rise of the 1% for the welfare of most citizens.

Here is Mike Konczal of the Roosevelt Institute on redistribution and its discontents:

A predominant Democratic view is that the economy is mostly fine; it’s just a matter of adjusting and correcting it to ensure everyone has access. Deeper, structural, changes are put to the side in favor of taxes, transfers, and behavioral nudges to help people out.

On trade, for example, the consistent Democratic narrative in 2016 was that we need to “compensate the losers” of trade. The phrasing alone tells us everything we need to know. Which voters want to be identified as losers? Democrats may mean something more abstract when they speak of “losers” in a globalized economy, but the language carries the connotation of personal blame.

But what role does individual agency play when global capital flows upend communities? And why are we treating the economy as a natural phenomenon — one whose consequences we simply must accept — when voters know it’s a series of laws, trade agreements, and businesses making decisions? If this is the best Democrats can offer, it’s not surprising workers aren’t interested.

It’s worth mentioning the Rewrite the Rules report from Roosevelt here, as well as the general idea of “predistribution” policy, coined by political scientist Jacob Hacker.

There is also a wave of backlash against tech titans’ endorsement of a universal basic income. Here is one good example from Helen Razer at Quartz:

Here’s the shameful secret not uttered in our favorite futurists’ TED-style presentations. The reason they adore UBI isn’t to do with their commitment to lift a growing underclass out of poverty; that’s just a bedtime story that helps the super-wealthy sleep. Instead, it’s more to permit spending on their goods by what remains of the American middle class. No one on a stagnant wage can currently buy the things that Musk—and the rest of Silicon Valley—wants to sell them. These billionaires champion a scheme whose prime result will be their profit.

The case for redistribution

Here is Matt Yglesias:

The solution to both facets of this problem is simple: taxes. Higher taxes on very high wages and higher taxes on investment income. Some of the revenue should go to the kind of earned income tax credit boost that Rep. Ro Khanna (D-CA) and Sen. Sherrod Brown (D-OH) have proposed, and some to create a universal child allowance of the sort that’s taken a huge bite out of poverty in foreign countries.

Here’s coverage of U.S. Representative Ro Khanna’s plan to dramatically increase the size of the earned income tax credit, in Vox and The Atlantic.

Here’s Dylan Matthews’ case for a basic income, and a study suggesting it would grow the economy.

This paper suggests redistribution from rich to poor has been the historical norm for the last hundred years, and this paper suggests it improves life satisfaction. That’s broadly consistent with the historical data presented here on “social spending” improving human well-being, which admittedly may not map well onto today’s debates between redistribution and “pre-distribution” policies.

State capitalism vs. the alternatives

Noah Smith has a nice column on state capitalism vs. democratic capitalism, and argues that it’s a battle of ideas akin to communism vs. capitalism. As he writes in the beginning:

The great experiment that Vladimir Lenin and Joseph Stalin began is over. And that experiment was a colossal failure. Market economies are necessary for getting rich.

The whole thing is in line with my post the other week about the case for capitalism and the mixed economy. Here’s a bit more from Smith:

Why are democratic countries turning to redistribution, while authoritarian powers seem to be reducing the role of government? One reason is that democracies tend to be richer, and wealthier nations simply have more money to spend on safety nets for their poorer citizens. It’s possible that as autocracies like China grow richer, their citizens will also demand generous welfare states — or even a transition to democracy.

But this isn’t written in stone. Many see Singapore as an authoritarian capitalist success story. The tiny nation is wealthier than almost any democratic nation, yet it remains a one-party state with low levels of government spending and a light regulatory touch. It seems possible that instead of following the path of the democratic-socialist nations, China and other post-communist countries will end up looking more like Singapore. They certainly seem to be aiming for something along those lines.

So although it’s too early to know for sure, it looks like a new division is replacing the old Cold War dichotomy of democratic capitalism versus authoritarian communism. In the new system, democratic-socialist countries will face off against authoritarian state-capitalist ones. It will be the Denmark model versus the Singapore model.

A generous welfare state is compatible with a dynamic, innovative economy

Two Brookings scholars have a great piece in Boston Review making the case that the safety net helps promote economic dynamism. And they make the case that the conventional wisdom is changing, even among some conservatives. And, sure enough, a few days later The New York Times ran an opinion piece by an entrepreneur advancing a similar argument, tied to the Trump tax proposals. (Silicon Valley entrepreneurs are actually quite open to redistribution.)

The Boston Review piece in particular is worth a read, and I’m grateful that they cite my writing on this subject. In light of their piece, I figured it’d be good to put a few things I’ve written on the subject in one place. Here’s the most extensive piece I’ve written on this, for The Atlantic. I have done a series of pieces for HBR: on health insurance, unemployment benefits, and college tuition. And I’ve posted a couple times here on the blog about others making similar arguments. Here’s one about Will Wilkinson, here’s one on Zuckerberg and a piece by Neil Irwin.

Between the Boston Review piece and my Atlantic piece, there are links to most of the relevant papers, and references to many of the key people making the argument.

Taxes and growth

Do lower taxes mean faster economic growth, as is so often claimed by conservatives? I mentioned this question in the context of corporate taxes recently:

Although in general low taxes do not necessarily increase growth, corporate taxes are considered “the most harmful type of tax for economic growth”, according to the OECDAnd researchhas found that decreases in the corporate tax rate spur investment, which in the U.S. has been surprisingly low in recent years.

I figured I’d post a few other resources here, along those lines. This is a review paper from Brookings on individual taxes and economic growth:

We find that, while there is no doubt that tax policy can influence economic choices, it is by no means obvious, on an ex ante basis, that tax rate cuts will ultimately lead to a larger economy in the long run. While rate cuts would raise the after-tax return to working, saving, and investing, they would also raise the after-tax income people receive from their current level of activities, which lessens their need to work, save, and invest. The first effect normally raises economic activity (through so-called substitution effects), while the second effect normally reduces it (through so-called income effects).

Here’s a Congressional Research Service report from 2012 that caused a lot of debate, and concluded that:

The results of the analysis suggest that changes over the past 65 years in the top marginal tax rate and the top capital gains tax rate do not appear correlated with economic growth. The reduction in the top tax rates appears to be uncorrelated with saving, investment, and productivity growth. The top tax rates appear to have little or no relation to the size of the economic pie.

You can read some coverage of that study here and here. And here’s an NPR fact check piece on the subject, which describes the Brookings piece mentioned above.

Here’s a survey of economists. They’re asked if, “A cut in federal income tax rates in the US right now would lead to higher GDP within five years than without the tax cut.” They’re divided between Yes and Uncertain, with only a few No’s.

And here’s a couple good pieces by Noah Smith. On individual taxes:

the best evidence that economists can muster shows that income taxes — i.e., what Republicans are always trying to cut — don’t hurt the economy very much. Microeconomic estimates of something called the Frisch elasticity of labor supply — or the amount that taxes discourage people from working — are very low. That means that income taxes do only a very little to discourage people from working. The one exception is tax cuts for the poor and working class, which really do seem to encourage more work effort. But for the upper-middle class and rich, who bear most of the tax burden and who are usually the prime beneficiaries of Republican tax cuts, the effect is very small.

And here he is on corporate taxes.

Eduardo Porter on taxes and growth, and an estimation of the optimal top tax bracket.

What’s the evidence on short-termism?

Here was my attempt to sum it all up with links a few months back, as the introduction to a Q&A with Steve Kaplan about a paper he had on the subject:

McKinsey’s Dominic Barton has made the case, as has BlackRock’s Larry Fink. Politicians like Hillary Clinton and Joe Biden have warned against short-termism, as have scholars at Brookings and the American Enterprise Institute. McKinsey has made its case empirically, finding evidence linking long-term management to superior financial performance. In 2015 Rotman’s Roger Martin reviewed the evidence on both sides here at HBR and explained why he believed short-termism is a problem.

But not everyone agrees.

Economist Larry Summers says, in response to the McKinsey data, that the jury’s still out. The Economist calls short-termism a “slippery idea” and a “distraction.” The New Yorker calls it a “myth.” And we’ve published many pieces here at HBR taking issue in one way or another with the standard short-termism critique.

In a recent paper, University of Chicago Booth economist Steven Kaplan makes his own case against worrying about short-termism.

Here’s a similar effort by Noah Smith, who makes the case that short-termism is, in fact, a problem:

Back in June, I reported on a research paper by Steven Kaplan of the University of Chicago’s Booth School of Business, saying that the threat of short-termism was either nonexistent or exaggerated. But I also argued that the reasons Kaplan gives have major caveats or are of questionable relevance.

Other research has shown important evidence on the negatives of short-termism. A 2010 paper by economists John Asker, Joan Farre-Mensa and Alexander Ljungqvist found that closely held companies tend to invest more than similar publicly listed companies, and also tend to be quicker to respond to new investment opportunities. And a 2007 paper by Rudiger Fahlenbrach found that companies run by founder-chief executive officers tend to invest more in both capital goods, and research and development — investments that are rewarded with higher stock prices over the long term.

The evidence that short-termism might be harmful continues to pile up. A 2014 paperby Stanford University’s Shai Bernstein finds that when companies go public and face pressure for quick results from investors, their best inventors tend to leave, and the ones who remain produce fewer patents. Though patenting is a poor measure of innovation at the industry-wide level (since one company’s patents can hinder innovation by other companies), it’s a good indicator of the effort a company is putting into research. Bernstein’s paper also shows that once companies go public, they plow less of their resources into far-sighted R&D investments.

Meanwhile, economists German Gutierrez and Thomas Philippon have a recent paper investigating the causes of low business investment. They find that the more public companies are owned by institutional investors, the less they tend to invest.

Posting this mostly so I have the links to all these bits of evidence together in one place.

More on how to think about economic models

I wrote recently about three different ways to think about economic models. Here are two more. John Gruber of MIT to his undergraduate micro students:

We’re going to be modeling individual and firm behavior. Now technically, as you know, a model is any description of the relationship between two or more economic variables.But the difference from your other courses– and I’m telling you right now, it’s going to be frustrating.I’m warning you in advance, is unlike the relationship between energy and mass, there is no law that tells you exactly how things relate.We’re going to build a series of models that is going to help us try to understand the way things relate. But this is not a real science. As much as we wish we were, we are not. We’re not a real science. We are a quasi-science, social science.What we’re trying to do with our models is make assumptions that negotiate the tension between, on the one hand, explaining real world phenomena, and on the other hand, being mathematically tractable.

Here’s the CORE economics textbook The Economy:


A good model has four attributes:

  • It is clear: It helps us better understand something important.
  • It predicts accurately: Its predictions are consistent with evidence.
  • It improves communication: It helps us to understand what we agree (and disagree) about.
  • It is useful: We can use it to find ways to improve how the economy works.

What’s the case for capitalism?

People often assert that socialism doesn’t work, or at least that capitalism is superior. Yet they seldom provide evidence for that claim. How do we know it’s true? If you know some economics, you can probably list some theoretical reasons why it could be true. Or perhaps you’ve read first-hand accounts of life under communism, and consider the question settled. But what empirical or historical evidence would you point to if someone asked you to make the case for capitalism relative to other systems of economic management?

I recently happened upon two attempts to answer this question, one from the excellent new intro econ textbook The Economy, and the other from some of the essays in the fantastic Cambridge History of Capitalism Volume 2.

The treatment in the textbook looks at East vs. West Germany:

The division of Germany at the end of the Second World War into two separate economic systems—centrally planned in the east, capitalist in the west—provided a natural experiment… The East German Communist Party forecast in 1958 that material wellbeing would exceed the level of West Germany by 1961. The failure of this prediction was one of the reasons the Berlin Wall separating East from West Germany was built in 1961. By the time the Berlin Wall fell in 1989, and East Germany abandoned central planning, its GDP per capita was less than half of that of capitalist West Germany…

Unlike some capitalist economies that had even lower per capita incomes in 1950, East Germany’s planned economy did not catch up to the world leaders, which included West Germany. By 1989, the Japanese economy (which had also suffered war damage) had, with its own particular combination of private property, markets, and firms, along with a strong government coordinating role, caught up to West Germany, and Spain had closed part of the gap.

We cannot conclude from the German natural experiment that capitalism always promotes rapid economic growth while central planning is a recipe for relative stagnation. Instead what we can infer is more limited: during the second half of the twentieth century, the divergence of economic institutions mattered for the livelihoods of the German people.

That’s a good start, and one of the nice things about The Economy is that it includes a lot of economic history relative to other introductory texts. But what about the overall record of capitalism? And what about in terms of human welfare, rather than just economic growth?

In an essay titled “Capitalism and human welfare” in the Cambridge History of Capitalism Volume II, Leandro Prados de la Escosura tackles this question, using a measure of human welfare combining economic growth (adjusted for inequality), life expectancy, and educational attainment. After providing and considering the data, he writes:

How do the capitalist and socialist systems compare? It has been frequently argued that it is at low levels of economic development when socialist societies have an advantage over capitalist ones in lifting human well-being and, in particular, its non-income dimensions. A glance at the former Soviet Union shows that substantial gains in human development were obtained between the 1920s and the 1960s, which resulted in an impressive catching-up to the OECD. Since the mid-1960s, however, this progress gave way to stagnation… A preliminary evaluation suggests that, but for Russia during the central decades of the twentieth century and Cuba, socialism has not delivered higher human development for developing countries than capitalism… The results presented in this chapter suggest that, despite its initial success as provider of ‘basic needs’, socialist experiences failed to sustain the momentum and, but for Cuba, stagnated and fell behind before the demise of communism. Furthermore, its suppression of agency and freedom prevented real achievements in human development.

Another essay in the same volume provides some detail, context, and theory to support the conclusion above. The essay is titled “Modern capitalism: enthusiasts, opponents, and reformers” by Jeffry Frieden and Ronald Rogowski. They write:

The centrally planned economies achieved rapid growth in the twenty years after World War II, as they drew underutilized resources into production. But Soviet-style planning had many limitations. As was true of the import-substituting economies, the Soviet bloc found that it increasingly needed imports — not only of food, but of technology and precision parts — that it lacked the hard currency to buy. Collectivized agriculture proved massively inefficient, forcing the formerly grain-exporting USSR to expend scarce foreign currency, year after year, on imported cereals. Recurrent campaigns to increase manufactured exports, particularly from the bloc’s most advanced economies (e.g., East Germany), brought little success. Only the bloc’s raw materials and a few artisanal products found ready purchasers in the West. The absence of incentives gave workers and managers little need to monitor quality, or to innovate either in the production process or with new products. Over time, the industrial plant fell farther and farther behind the technological and quality criteria prevailing in the West, and by the 1980s growth had slowed dramatically. With Western Europe within easy reach of people in the Soviet bloc’s central and eastern European nations, it was easy for citizens to see the relative failure of the system.

Together, it seems reasonable to conclude that the conventional wisdom is roughly accurate. But each source provides reasons for humility. The success of capitalism over socialism may be apparent, broadly and in aggregate. But the historical record contains various partial exceptions. We ought to be cautious about overly sweeping generalizations.

For one thing, the chapter mentioned above on capitalism and human welfare provides an important reminder that the victory of “capitalism” is really a victory for the mixed economy. The author documents “a positive non-linear association between the expansion of social protection and the improvement in human development… Small changes in social transfers are associated with large increases in human development. Then, as we move to the right, we observe that increases in social transfers are associated with smaller, but still positive, increases in human development. As social transfers reach 25 percent of GDP the curve tends to flatten, suggesting a reversal for levels above 30 percent.”


But does social spending actually cause improvements in human welfare? Or is it merely correlation? The answer comes from yet another essay in the volume, “Private welfare and the welfare state” by Peter H. Lindert. Yes, he says, at least historically:

The modern rise of public social spending probably brought considerable gains in efficiency, GDP, and the larger concepts of human welfare quantified in Chapter 15 in this volume by Leandro Prados de la Escosura. These investments in humans were blocked for millennia by weak and rapacious governments, and by a concentration of political power that rejected universal public schooling, family assistance, and public health insurance…

Yet since about the 1960s, the expansion of public social programs has probably stopped reaping efficiency gains, due to what journalists would call ‘mission creep.’ Several countries, most notably Japan, the United States, Italy, and Greece, have drifted away from their original mission of investing in the young, while at the same time maintaining intergenerational transfers in favor of the elderly. This drift did not bring any obvious net loss of GDP in the late twentieth century, but further population aging in the twenty-first will force reforms that hold support for the elderly within sustainable steady-state limits.

To summarize, then: We can be reasonably confident that capitalism is broadly superior to socialism and state planning, with the caveat that by capitalism we mean a mix of free but regulated markets and a significant welfare state. The historical evidence strongly suggests that social transfers improve human welfare. And some of those transfers — specifically transfers to the young — also improve economic efficiency and spur economic growth.

The linear regression economy

This is a point I meant to blog about, but a new piece from Tech Review led me to more or less sum it up in tweets:

A related point: models are not the constraint for most data science projects. My thinking in all of this is informed by Kalyan Veeramachaneni of MIT, who’s written about some of these issues here.

Amazon, Sears, and market power

Derek Thompson has a nice piece at The Atlantic about the uncanny similarities between Amazon and Sears (a point Benedict Evans has also made):

It’s remarkable how Sears’s rise anticipates Amazon’s. The growth of both companies was the result of a focus on operations efficiency, low prices, and a keen eye on the future of American demographics.

Might this comparison say anything about recent concerns over Amazon’s market power?

Well, you might think the story of Sears illustrates the problems of market power — a big company extends its tentacles, competing with much smaller ones, and controls an extremely large part of the market for lots of different stuff.

But here’s Robert Gordon on what Sears replaced:

In the America of 1870, local merchants had local monopolies, and their customers had little ability to compare prices. Thus there was constant tension between the rural customers and the merchants, for there were few guidelines for judging whether a price was fair… In this environment, we can understand why mail-order catalogs were such a salvation for rural customers.

The Sears story, then, is about the creation of a big, powerful company that, yes, probably gained all sorts of market power via its scale. But it replaced lots of little geographic monopolies — small firms that each had substantial market power and used it to extract rents from customers. The change wasn’t about moving from a competitive environment to a less competitive one. It was about changing from one form of imperfect competition to another one.

Could something similar be true of Amazon? Who knows — all this is speculative at best. But the Sears comparison clearly helps understand Amazon in other ways, so perhaps it’s at least worth thinking about.