Saturday, July 4, 2020

"My Country, Right or Wrong": No Patriot Would Say It

In the 21st century, I'm not sure how many Americans would ever actually say "my country, right or wrong." After all, it's not "countries" that are right or wrong, but actions of governments and people, and it seems ingrained in the American character (and thankfully so!) that criticizing one's government is not only acceptable, but often expected.

But perhaps the ultimate put-down for that point of view came from G.K. Chesterton, In a 1901 collection of essays, The Defendant, he includes an essay called "A Defense of Patriotism." Chesterton writes:
My country, right or wrong,' is a thing that no patriot would think of saying except in a desperate case. It is like saying, 'My mother, drunk or sober.' No doubt if a decent man's mother took to drink he would share her troubles to the last; but to talk as if he would be in a state of gay indifference as to whether his mother took to drink or not is certainly not the language of men who know the great mystery.
If faced with a situation where the government of a country has done something terribly wrong, and considering whether to betray the country as a result, even a patriot might defend an ultimate loyalty to the nation by saying "My country, right or wrong." But it's a sentiment that would only come up when the "wrong" was deeply and profoundly wrong, and the immediate options were grim. It's a statement that would arise only from a spirit torn with near-despair and great humility, and would would not be made in a spirit of pride, or even defiance.

A Refresher on Humans, Angels, and Government

One can loosely divide American preferences about the role of government into two groups: those who think that because people and markets are not angels, more government actions are needed, and those who think that because people and government is not full of angels either, fewer government actions are needed. A classic statement behind this dilemma comes from James Madison in the Federalist Papers #51. In discussing the design of the US government, he wrote:
"Ambition must be made to counteract ambition. The interest of the man must be connected with the constitutional rights of the place. It may be a reflection on human nature, that such devices should be necessary to control the abuses of government. But what is government itself, but the greatest of all reflections on human nature? If men were angels, no government would be necessary. If angels were to govern men, neither external nor internal controls on government would be necessary. In framing a government which is to be administered by men over men, the great difficulty lies in this: you must first enable the government to control the governed; and in the next place oblige it to control itself. A dependence on the people is, no doubt, the primary control on the government; but experience has taught mankind the necessity of auxiliary precautions."
In arguing for the government to control itself, Madison laid out a system with a separation of powers with two different legislative houses; an executive branch headed by a separately elected president; and an appointed-but-subject-to-confirmation judicial branch. The idea wasn't new to Madison, of course. Arguments for the separation of legislative and executive are made in John Locke's (1690) Second Treatise on Government, although Locke tended to view the power of judging, via magistrates, as an offshoot of legislative power. Montesquieu is often credited with the the first explicit division of governing powers into legislative, executive and judiciary in his 1748 book The Spirit of the Laws. For example, Montesquieu writes (in translation) in Book XI, Chapter 6:
When the legislative and executive powers are united in the same person, or in the same body of magistrates, there can be no liberty; because apprehensions may arise, lest the same monarch or senate should enact tyrannical laws, to execute them in a tyrannical manner. Again, there is no liberty if the judiciary power be not separated from the legislative and executive. Were it joined with the legislative, the life and liberty of the subject would be exposed to arbitrary control; for the judge would be then the legislator. Were it joined to the executive power, the judge might behave with violence and oppression.
Economics is not willing to assume that angels are taking human form--not among business executives, workers, consumers, taxpayers, voters, nor among who are elected, appointed, or hired to government posts. It can feel frustrating when society and government seems to struggle against itself and may even become unable to act, with conflicts between branches and their differentiated responsibilities. But the alternative of a government in which acts without such constraints is, at least to me, a considerably less attractive prospect.

Friday, July 3, 2020

US Immigration: Attitudes and Workplace

Each year in June, the Gallup organization does a national poll on attitudes toward immigration. Here's the answer to a basic question: should immigration be kept at its present level, increased, or decreased? 
As you can see, from the 1960s up through the mid-1990s, the share who wanted to increase immigration was quite low at about 7%. In 1995, 65% wanted to decrease immigration. 
Immigration1
But opinions on immigration have shifted dramatically in the last 25 years and perhaps especially since around 2010. The "present level" group has often been roughly one-third of the total over time, with some fluctuations. But the share wanting "increase" has been rising over the last 25 years,and in the early June survey it came in higher than the share wanting "decrease." 

One reason for the shift, I suspect, is that the foreign-born share of the US workforce has been rising during this period. Moreover, the foreign-born workers as a group have become more likely to have college degrees.  One additional factor is that the rise number of immigrants without legal status flattened out back in the 2007-2009 recession, and the number of people in this category hasn't been rising since then. A recent Congressional Budget Office report provides a data-based look at "The Employment of Foreign-Born People" (June 2020) wrote:   
The foreign-born population of the United States was 46 million in 2018, representing 1 out of every 7 people. Twenty years earlier, in 1998, 1 out of every 10 people had been born abroad. Over the past two decades, the foreign-born population increased by 61 percent, outpacing the 13 percent increase in the native-born population. A key component of that growth was the increase in the number of foreign-born people with legal status; by CBO’s estimate, that number grew from 20 million, or 72 percent of foreign-born people, in 1998 to 35 million, or 76 percent of foreign-born people, in 2018. At least 90 percent of foreign-born people with legal status were naturalized citizens or lawful permanent residents, according to data from the Department of Homeland Security; the rest had temporary legal status usually associated with study or work. The numbers of foreign-born people both with and without legal status grew between 1998 and 2007, according to CBO’s estimates, but most of the growth thereafter was among people with legal status.
Moreover, the CBO report shows an ongoing shift toward higher education levels for the foreign-born in the last couple of decades. 
This shift has been widely noted. As one example, a National Science Foundation report pointed out earlier this year
Foreign-born workers—ranging from long-term U.S. residents with strong roots in the United States to more recent immigrants—account for 30% of workers in S&E occupations. The number and proportion of the S&E workforce that are foreign born has grown. In many of the broad S&E occupational categories, the higher the degree level, the greater the proportion of the workforce who are foreign born. More than one-half of doctorate holders in engineering and in computer science and mathematics occupations are foreign born.
I'm sure that part of the shift toward encouraging more immigration is less about the issue itself, and is just politics: in this case, a pushback against President Trump's anti-immigration rhetoric. For example, Democrats rather abruptly began to increase their support of free trade after Trump took office threatening protectionism early in 2017. But these underlying shifts in attitudes toward immigration, along with shifts in numbers and skill levels, didn't just start in the last few years: they have been evolving over time. 

Thursday, July 2, 2020

Religion and Life Outcomes: Looking for Causal Effects

The pervasiveness of US religious belief has seemed to decline in the last 30 years or so, by a variety of measures. Daniel Hungerman investigates "Religious Institutions and Economic Wellbeing" in the most recent issue of Future of Children (Spring 2020, pp. 9-28). As a starting point, here's a figures from Hungerman showing the fall in US religious belief over time (using data from the General Social Survey). 
The decline in religiosity seems likely to continue, because the shift is especially large among younger age groups. 
Or here's a figure showing that the share of charitable giving going to religious organizations was about half back in the 1980s, but it now down to about 30%. 

Why this shift started happening around 1990s isn't altogether clear. The sex scandals afflicting the Catholic Church, for example, are not especially prominent in the broad public eye until the early 2000s. There's some nondefinitive evidence that increasing levels of education may tend to reduce religious belief. There's also a sense that in  many people's minds, religion has become more entangled in politics, which has led some people to react by drawing back from religious belief.  

But Hungerman's main focus is about the effects of this change. It's well-known that religious people often typically report being happier, and are more likely to vote, less likely to use drugs or commit crimes, and so on. But correlation isn't causation. Ideally, one might want to do an experiment where some people randomly become committed to religion while an otherwise identical comparison group does not do so, but this particular experiment seems impractical. Indeed, it seems plausible that those who choose to participate in religion may differ in some underlying way from those who do not. So how can a researcher try to find evidence, one way or another, of causal effects? 

I'll just say up front that there's no perfect way to find a pure causal effect from religion to economic outcomes. Instead, like a lot of social science issues, one instead tries to tackle the question from a variety of angles in specific research studies, and then see what overall pattern starts to emerge. Thus, the array of examples also gives a sense of how the minds of social scientists work. For example, here are some of the examples and approaches discussed by Hungerman:  

Matching people who are equivalent in the non-religious variables we can observe
Dehejia, DeLeire, and Luttmer examine whether religious individuals’ consumption and self-reported wellbeing appear to be relatively less sensitive to income shocks—that is, whether religion helps “insure” people against negative shocks. ... [T]he authors use a variety of methods, such as applying a procedure that matches each religious person in a sample to an observationally similar nonreligious person, so that the final data sample contains a similar distribution of observable characteristics across religious and nonreligious individuals. They find that religiosity does indeed insure against negative shocks.
Looking at whether more of your neighbors share your religious tradition
Gruber proposes a creative strategy: using variation in the ethnic composition of one’s community to study the impact of religion. Put simply, an American of Italian ancestry may not make much of a distinction between living in a neighborhood full of Swedish individuals versus a neighborhood full of Polish individuals—except that the latter group, like Italians, are Catholic. If living side-by-side with ethnicities that share your religious tradition makes you more religious, but otherwise doesn’t affect your wellbeing, than we can use ethnic composition to learn about the causal effects of religion.
Gruber finds, again, that religiosity leads to better outcomes for a number of economic indicators.
Peer groups in schools
Especially noteworthy is a study by the economists Jane Fruehwirth, Sriya Iyer, and Anwen Zhang. In an approach similar to Gruber’s, they exploit variation in the religiosity of peers across cohorts within a school to identify how religion influences mental health in a sample of US adolescents. They find that religion plays an important causal role in promoting mental health.
A lottery for attending the Haj
David Clingingsmith, Asim Khwaja, and Michael Kremer. They look at the effects of attending the Hajj—the pilgrimage to Mecca that Muslims are expected to make at least once during their lifetime. To study how attending the Hajj affects people’s values, Clingingsmith, Khwaja, and Kremer use a Pakistani lottery that allocates Hajj visas; they find that participation in the Hajj leads to greater acceptance of female education and employment. More generally, Hajj lottery winners show both increased Islamic observance and greater belief in equality and harmony among all religions.
Distance from Wittenberg
The great social scientist Max Weber famously considered whether a Protestant ethic for work might drive the difference between economic wellbeing in Protestant and Catholic communities. Becker and Woessmann take up this association in several steps. First, they put it to a careful test in historic Prussia, exploiting the fact that Protestantism expanded from its birthplace in Wittenberg (a previously unimportant town) in a pattern akin to concentric circles. Moving away from Wittenberg, you encounter all sorts of terrain and all types of communities—but places farther from Wittenberg are less likely to be Protestant, all else equal. Becker and Woessmann then confirm that distance from Wittenberg appears unrelated to various controls (such as the presence of schools in the 1500s, before the reformation), but centuries later it does predict income and economic circumstance—being closer to Wittenberg (and therefore more Protestant) is better for economic wellbeing. This suggests that the link between GDP and Protestant affiliation is more than a simple association. Does this mean Weber was right? Not quite. The final step of Becker and Woessmann’s study shows that variation in literacy can largely explain the economic gains of Protestantism. It appears that the Protestant emphasis that everyone should be able to read the Bible (and thus be able to read), rather than a “noncognitive” work ethic, can explain why Protestant societies had higher economic productivity.
Interaction of religion with laws about alcohol and gambling
Looking at the United States, Jonathan Gruber and I investigated this by looking at the repeal of “blue laws” that restrict economic activity on a certain day of the week (often Sunday).40 Most recent blue laws are narrow in focus—for example, alcohol can’t be sold at grocery stores before noon on Sundays. But not that long ago, many states had strong blue laws that prohibited most Sunday economic activity. A Supreme Court ruling in 1961 provided a test by which these laws could be repealed, and many were consequently undone. Gruber and I show that when such laws are undone, religiosity declines, and that risky behavior such as heavy drinking increases— but the increases are driven by those who report having been religious before the repeal occurred.  ... Religious rules appear to be effective in discouraging heavy drinking and gambling. The results [from another study] often indicate that the most religious individuals are those who are likeliest to substitute: it’s the most religious groups whose religious giving declines when casinos open or when commerce is allowed on Sundays, and it’s the most religious individuals who are likely to start drinking heavily when the legal drinking age changes.
Figuring out causal connections from religion to life and economic outcomes is a challenging research project. But the weight of the evidence Hungerman discusses--only a bit of which I've mentioned here--is that those who end up being exposed to religion do seem to experience measurable benefits, which in a broad sense take the form of bolstering a person's confidence and determination in following a path of learning, saving, and work--and avoiding being derailed by overindulgence in counterproductive habits. 

I'll also append here the full Table of Contents for this issue of Future of Children. It seems to have an even higher-than-usual proportion of interesting essays, and I may post some additional commentary about other essays in the next week or so. 

Wednesday, July 1, 2020

The Productivity That Didn't Happen

The bad news is that US productivity growth has been slow for the last 15 years, and in facts for 30 of the last 40 years. But at least other high-income nations are doing worse. Emily Moss, Ryan Nunn, and Jay Shambaugh provide a nice readable overview of productivity fact patterns, along with possible causes and solutions, in "The Slowdown in Productivity Growth and Policies That Can Restore It" (June 2020, Hamilton Project at Brookings). 

It's conventional to divide US productivity growth since 1948 into four periods, summarized in this figure. There's the reasonably rapid post-World War II productivity growth from 1948 to 1973, the slowdown from 1973-1995, a 10-year resurgence from 1995-2004, and the return-to-slowdown since then. 
The US situation doesn't look good, but in fact, we're going better than other high-income countries. The fact that the productivity slowdown encompasses all the high-income countries has an important implication: it suggests that at least some of the most important causes are not specific to US economic policy or indeed to the policies of any one country, but instead must be causes that would apply broadly across all high-income countries. 

Don't forget that the numbers on the figures above are annual rates. For example, US labor productivity drops from an annual rate of 3.1% from 1995-2004 to an annual rate of 1.4% from 2004-2018. This annual change accumulates over time.  To get a feeling for the importance of this accumulation,  let's say that labor productivity had continued to rise by 3% per year since 2004. The extra growth could have led to gradually higher incomes every year. As a result of labor productivity growing 1.6% per year faster over the last 16 years, the total US economy in 2020 would be 29% larger. 

Given that US GDP will be about $22 billion this year, 29% works out to $6.4 trillion larger. For perspective, $6.4 trillion works out to roughly an extra $20,000 in 2020 for every US citizen, including adults and children. This is not a one-time boost, but a permanent and ongoing rise. For people, for government, for social problems, for the environment--every problem is a little easier to solve when the financial constraints are relaxed. But because productivity growth slowed down in 2004, those resources never came into existence. 

The reasons for these rises and falls in productivity are largely mysterious to economists. The problem is not a lack of hypotheses, but rather a sense that when you offer lots of possible explanations, perhaps you aren't all that sure about any of them. For example, if the more recent productivity slowdown had kicked in after the Great Recession, one could come up with a hypothesis related to the Great Recession--but it clearly started well before that. 

One possible explanation is that this is mostly a mismeasurement problem. The argument here is that GDP doesn't capture the economic gains of new technology, so the productivity gains don't show up in official statistics. There's no doubt that the official statistics are imperfect, but but the question here is whether they somehow became more imperfect circa 2004; that is, they captured the rise in productivity growth and the web pretty well for a decade, but then stopped doing so. There isn't much evidence to support that belief. Moreover, the decline in US productivity after the 1995-2004 burst has been quite broad across industries. To put it another way, it doesn't seem as if productivity is only down in certain harder-to-measure industries. 

In addition, while it would certainly be nice to believe that our standard of living is rising in all sorts of ways not measured by actual dollars, our household bills and mortgage debt and taxes and government borrowing need to be paid with actual money, not just with a nebulous feeling of being better off. 

While the struggle to explain the timing and breadth of the changes in productivity goes on, some of the intellectual energy has instead shifted toward thinking about what might help reverse the pattern, regardless of its underlying cause. I sometimes like to say that there's the basic formula for productivity is well-understood: it's a combination of human capital, physical capital, and new technology, interacting in an economic environment with incentives for innovation. Here are a few thoughts from Moss, Nunn, and Shambaugh on these issues, with more in the actual report: 

On the issue of human capital, America's rise in average education level has slowed down, and the aging population means slower growth in the prime-age labor force. They write: 

For cohorts born from 1876 to 1951, average educational attainment rapidly increased by 0.8 years per decade, with successive generations receiving about two additional years of education relative to their predecessors. The pace of this increase has now slowed: cohorts born from 1951 to 1987 have added only about 0.3 years per decade ...
Slower growth in the prime-age labor force tends to coincide with slower growth in productivity, perhaps because of a reduction in available managerial talent (Feyrer 2007, 2011) or the rate of business formation (Karahan, Pugsley, and ┼×ahin 2019). The aging of the workforce can also place downward pressure on productivity growth by making it more difficult to implement new innovations and processes ...
When it comes to private investment, firms don't seem to be doing a lot more of it. Here's a figure showing investment by firms in the specific area of information processing equipment and software. It's probably not a coincidence that the sharp rise starting in about 1990 was followed by rising productivity a few years later, and conversely that the drop in 2000 was also followed by lower productivity a few years later. 
One my my hobby-horses on this blog is the need to increase research and development spending. As the figure shows, total R&D as a share of GDP hasn't risen for decades. But what is changing is that the federal share of R&D which tends to focus on basic research and thus on potential breakthrough innovations is falling, while the business share of R&D which is more likely to focus on near-term development of products is rising. 

When it comes to increasing productivity, there are a number of other areas that deserve attention, many of which involve trying to strike a better balance: Can we figure out ways to provide inventors with a return that also encourage follow-up inventions by others? Can we figure out ways to encourage competition and limit anticompetitive behavior? Could adjusting rules related to occupational licensing or residential building help labor to become more productive? Could investments in infrastructure for transportation, energy, and communications help labor and industries be more productive?

The only way for the average person in a nation to consume more in the long-run is for that average worker to have higher productivity in the long-run. Yes, for a time it's possible to raise taxes on the rich and transfer to others. It's possible for the government and firms and people to borrow money for a time and raise consumption in the present, too. Redistribution and borrowing are useful for specific circumstances, and for certain times and places, but by themselves, they can't continually raise consumption for the average person. Only rising productivity can do that. 

Saturday, June 27, 2020

We Pay Plasma Donors, Don't We?

The quantity of blood plasma being used in medical therapies was already rising at 6-10% a year. Now there is some evidence that blood plasma products may be useful in treating COVID-19. Peter Jaworski writes (footnotes omitted: 
Transfusing the blood plasma of those who have recovered from Covid-19, called convalescent plasma, appears to help against the novel coronavirus. The very same treatment has been used successfully as early as the 1918 Spanish Flu epidemic, and more recently with more similar infections such as MERS and the first SARS. The UK’s National Health Service has already begun trials of transfusing blood plasma of recovered Covid-19 patients. Convalescent plasma may also soon find another use in the form of a hyperimmune globulin. Hyperimmunes are plasma-derived medicinal products, or plasma therapies, made from isolated specific antibodies, in this case against SARS-CoV-2.
But here's the economic problem. The only way that quantity supplied has been keeping up with the rise in demand for blood plasma is through the efforts of the few countries in the world--primarily the United States--that allow blood donors to be paid. Thus, Jaworski has written "Bloody Well Pay Them: The Case for Voluntary Remunerated Plasma Collections" (2020, published by the Adam Smith Institute and the Niskanen Center). Here's how he described the global supply of blood plasma: 
The United States is responsible for 70% of the global supply of plasma. Along with the other countries that permit a form of payment for plasma donations (including Germany, Austria, Hungary, and Czechia), they together account for nearly 90% of the total supply.... Countries that exclusively use non-remunerated plasma collections for domestic plasma collections rely increasingly on the United States. The United States currently supplies approximately 70% of the global need, including about two-fifths of Europe’s needs, nearly all of the UK’s, over four-fifths of Canada’s, over half of Australia’s, and around 12% of New Zealand’s needs for plasma therapies. To put this into perspective, at present, 5% of the world’s population is responsible for more than half of all the plasma collected in the world. ... 

The global plasma supply situation has resulted in a staggering volume of plasma and plasma therapy exports from the United States. According to an estimate by The Economist, exports of plasma and plasma therapies represented approximately 1.6% of total exports by GDP, or U.S.$26 billion, in 2018, up from US$23.6 billion in 2016. To put this figure into perspective, that is more than exports of steel and aluminum, or the 11th most valuable export by value. Current projections, meanwhile, suggest that this figure will rise to U.S.$44.3  billion by 2023.
When it comes to blood plasma, many nations of the world have ended up in a position where paying for plasma is viewed as a moral wrong within their own country, but purchasing plasma from paid donors in the US and elsewhere is an everyday acceptable medical practice. This situation was already on shaky ground before COVID-19; if it turns out that blood plasma and its products are a useful part of a treatment for the coronavirus, then the quantity demanded of plasma will far outstrip available supply.  Jaworski writes: 
It is no longer reasonable, given the evidence, to continue to insist that non-remunerated plasma collections are able to meet domestic needs, never mind the global needs. Non-remunerated plasma collections have failed everywhere to secure a sufficient supply. There is now not a single jurisdiction anywhere in the world that meets its needs exclusively based on non-remunerated plasma collections.
In the United States, blood itself is usually collected only from volunteers, while blood plasma is from paid contributors. If it seems important to maintain this distinction, medical technology allows one to do so. Plasma can be either be collected as part of regular blood donations (the blood is spun in a centrifuge to separate red and white blood cells from the plasma), but plasma can also be donated separately. As Jaworski describes it: 
“[S]ource plasma” ... separates the plasma from the other blood components while the donor is in her chair, allowing the centre to capture only plasma while returning the remaining blood components to the donor. This process, called “plasmapheresis,” permits a donor to not only donate much more frequently but also to donate a much higher volume of plasma as compared to the plasma contained in a whole blood donation.
The question of blood plasma is not the only way in which COVID-19 is going to challenge some of our social intuitions about markets and medical discoveries. For example, consider the problem of testing a new vaccine. You pick a sample group of, say, 10,000 people. Randomly vaccinate half of them, while others get a placebo. Then have those people go about their business, waiting for some of them to get infected. But an obvious problem arises. When researchers at Oxford talk about testing a vaccine in this way, they raise the concern that of those 10,000 original people, maybe only 50 will end up being randomly exposed, maybe less. It's possible that so few people in the study group will end up being exposed that the test of the vaccine will not be statistically valid. And all of this takes time. 

In a purely practical sense, the fastest and most effective way to test a potential COVID-19 vaccine is to take a group of healthy people, vaccinate half of them at random, and then infect the entire group with COVID-19. One could either call for unpaid volunteers for such a project, or one could offer payment to a group of relatively healthy people. The payment could of course include compensation for lost wages, health care costs, living expenses, even life insurance. Instead of hoping for 50 people to be randomly exposed, make the exposure happen. For an argument along these lines, Ilya Somin makes "The Moral Case for Testing Coronavirus Vaccines through `Challenge Trials' on Paid Healthy Volunteers" (Reason website, May 27, 2020). 

It's of course true that paying for blood plasma or paying for vaccine trial volunteers raises some genuine moral issues. I won't pretend otherwise. But conversely, not paying also raises some genuine moral issues for opponents of such payments. 

For some earlier posts on paying for plasma and blood, see: 

Friday, June 26, 2020

Updating GDP: Human Capital, Nonmarket Work, Inequality, Health Care, and More

No one who knows anything about gross domestic product, including economists and government statisticians, thinks it is a broad measure of social well-being. Moreover, economists at the Bureau of Economic Analysis don't want the job of coming up with a broad measure, either.  In the June 2020 issue of the Survey of Current Business, J. Steven Landefeld, Shaunda Villones, and Alyssa Holdren provide an essay on "GDP and Beyond: Priorities and Plans." They write: "BEA economists use market prices and volumes to measure GDP and other economic aggregates and have no special expertise in deciding the appropriate weights to combine indicators such as education and life expectancy with gross national income per capita." 

It seems sensible to me to have the government focused on collecting a wide array of statistics across a full range of economic and social indicators, and then having outsiders debate the ultimate meaning of "better off." But as Landefeld, Villones and Holdren point point out, along with a set of outside commenters, there are plenty of challenges even when just sticking to what can be measure or at least estimated based on prices and quantities. Here are some comments and thoughts on what BEA lists as its top priorities moving forward. 

Human Capital

We have known for a long time that investments in education, skills, and experience lead to greater "human capital" which pays off for the economy as a whole. But when we discuss "investment" as a part of GDP, we remain focused on purchases by firms of plant, equipment, and software along with intangible investment from corporate R&D. Landefeld, Villones and Holdren mention one striking estimate that one estimate comparing the accumulated stock of human capital with the accumulated stock of physical capital, the stock of human capital may be worth 16 times as much. 
But there's a real challenge here in measuring human capital investment. It can be estimated in various ways, which don't agree. In addition, it seems  plausible me that human capital in the form of education is a mixture of a consumption good and an investment good, which complicates matters further. Here's one of the outside comments from Lisa Lynch:
There are three main approaches for measuring human capital investment for the purpose of national accounts: Kendrick’s (1976) Cost-based approach; the Lifetime income approach as developed by Jorgenson and Fraumeni (1989 and on); and the Indicators approach as detailed by the OECD 2011 (and updated every two years since), and Barro and Lee 2013. Measuring investment in human capital based on costs typically includes spending on schools, employee training costs, opportunity cost of time acquiring human capital, and a range of expenditures on others items such as libraries, radio, TV, books and other items having human capital value. The Jorgenson-Fraumeni (J-F) Lifetime income stream focuses on the present value of the return to formal education only. Finally, the Indicators approach pulls together a range of metrics such as adult literacy, school enrollments rate, average years of schooling, and the percentage of highly qualified workers to capture differences across countries and time in human capital investment.

In principal the Cost-based and Lifetime income approaches should produce values equal to each other. In practice they do not. The Lifetime-income approach produces estimates of investments in human capital 6 to almost 10 times greater than the Cost-Based approach. ... 
Household production

A standard concern about GDP is that it doesn't measure household production that is not sold on the market, and standard estimates are that including estimates of the value of household production would raise GDP by about one-quarter.  But unpaid economic output purely for household production is only the start of the issues here. What about tasks like elder care, volunteering, and pumping your own gas or bagging your own groceries? Here's another comment from Lisa Lynch
While there has been significant work done by the BEA to develop a satellite account for household production I would urge the BEA to add additional non-labor market activities that take place outside the home but meet the threshold of “Would someone pay another person (a “third person” from outside the home) to perform the activity?” The first such activity is elder care. We know from the American Time Use Survey (2014–15) that approximately 16.2% of the U.S. population provides eldercare—unpaid care for someone over the age of 65 with a condition related to aging. Almost all of this care takes place outside the home and on an “average” day, 26 percent of unpaid eldercare providers spend an average of 3 hours in eldercare activities. With an aging population this is a growing dimension of household production that should receive increasing attention in household satellite accounts.

A second area of non-labor market work that is not captured in our satellite accounts is volunteering. From the 2015 American Time Use Survey we learn that 9% of those over age 64 volunteer on an average day. For all those volunteers over age 25 they spent an average of 2.25 hours in this activity. Examples of volunteering include administrative and support activities, social service and care activities, and indoor and outdoor maintenance, building, and clean‐up activities. While not all aspects of volunteering may meet the standard of paying another person for this work, much of it would.

A third area of non-labor market work includes the “free labor” facilitated by IT. Examples of this include ATM bank transactions, self-service work of pumping gas and bagging groceries, online airline ticket purchase and check-in, “self-service” baggage tagging/drop, self-service keyless check-in and checkout at hotels, and ordering, paying and self-pickup of meals. None of these economic activities are captured in our national accounts today even though we still have employees who are paid to do this work.
Distribution of Income

When thinking about growth of the overall economy as measured by GDP, and evaluating its costs and benefits, it seems useful to know what distribution of income is happening as well. But measuring income distribution in a way that links up directly to GDP is harder than it may sound at first. It's common for measures of income distribution to focus on personal income, using either survey data or administrative data (like tax returns). But such measures are mostly limited to personal income, which is only about half of GDP. Could measures of distribution be linked to overall GDP? In his comment, Angus Deaton sounds a cautionary note about this agenda: 
Piketty, Saez and Zucman (PSZ) have done a great service by calculating a set of distributionally disaggregated national accounts for the United States. The basic idea is irresistible. Yet these first attempts have raised many serious difficulties that were not apparent at first. Most immediately, only about half of national income appears on individual tax returns. Allocating from tax returns is hard enough, because tax units are neither individuals nor households, but allocating the other half of national income is an immensely more difficult task, requiring assumptions that are rarely well supported by evidence, and often seem arbitrary. Because distribution is such a controversial topic, these assumptions leave plenty of scope for politically-biased challenges, in which each commentator can choose their own alternatives and get almost any result they choose, inequality is increasing, inequality is not increasing, and everything in between. It is surely not good practice for statistical offices, as opposed to researchers, to have to make such deeply controversial choices. My own preference would be to give up on the bigger task of measuring the distribution of national income, and to stick to the feasible, but still difficult, challenge of allocating personal income, both before and after taxes and benefits.
Health Care

The usual idea of measuring GDP is to look at a product, and then to measure the price times the quantity produced. For products like steel or oil or cars or pizzas or haircuts, this works pretty well. But health care is an ever-evolving mixture of inputs, and the specific output that is being purchases is hard to measure.  Thus, a standard way of measuring "output" in health care has been to look at total spending on healthcare--which is of course a measure of inputs, not of outputs like reducing cases or morbidity or deaths from diabetes or cancer. Health care expenditures are a huge and growing, headed for one-fifth of all GDP. But there is often some skepticism over whether the rise in health care spending as a share of GDP represents a rise in outputs--like improved health. In his comment, Angus Deaton puts the point with some force: 
The American healthcare system poses one of the most serious challenges to national income measurement, and plays into its well-known weaknesses. Like other services, it is measured, not by its output in terms of its contribution to health, but by its inputs, such as the number of procedures, doctor visits, or prescriptions sold. I do not know how to do better than this, but I do know that these numbers, currently about 18 percent of GDP, vastly overstate any imaginable output measure. Americans have lower life expectancy and higher morbidity than do other rich countries, who spend much less. To take a concrete example, Switzerland, which has the world’s second most expensive healthcare system, spends only 12 percent of its GDP. So one measure of the value of the output of American healthcare is 12 percent of American GDP, which would mean that the sector has negative value added of a trillion dollars a year; to balance the accounts, this trillion dollars would show up as poll tax on consumers, $8,700 per head, which is being transferred as a tribute, or ransom, from consumers to healthcare providers. Of course, I am not suggesting that the BEA adopt this calculation, but it illustrates the dangers of not having a measure of output and of accepting valuation at cost.
The paper and comments raise a variety of other issues.  Those who want still more detail might head for a "Symposium: Are Measures of Economic Growth Biased?" in the Spring 2017 issue of the Journal of Economic Perspectives (where I work as Managing Editor):