GDP and Beyond

Summaries from the 2020 Annual Meeting of the American Economic Association

Angus Deaton is a Professor of Economics and International Affairs, Emeritus, at Princeton University and a Presidential Professor of Economics at the University of Southern California.

The views expressed in this paper are those of the author and do not necessarily represent the U.S. Bureau of Economic Analysis or the U.S. Department of Commerce.

GDP has recently faced unusually intense criticism with some commentators calling for it to be replaced by a more direct measure of wellbeing based on self-reports. One problem is not the concept itself, but the way it is used, and that too much is expected of it. The use of GDP as the headline measure of economic progress is particularly unfortunate, and has become more so as economies have changed and become more globally connected. I shall say something about this first, and then about how we should extend national accounts to handle distributional issues, and finally about the problem of measuring healthcare. I should also note that while I am very much in favor of the collection of self-reported measures of wellbeing, including both evaluations of wellbeing and reports of feelings, I do not think it makes sense to think of them as replacements for GDP or other measures in the national accounts. Extensive work has proved that these measures are useful, that they sometimes capture important aspects of life that are not otherwise measured, and that much can be learned from comparing them with other, more familiar, measures. I think it is unfortunate that self-reports of wellbeing are not regularly collected somewhere in the American statistical system.

Much of the difficulty with GDP comes from thinking about it as a measure of material wellbeing. Each letter G, D, and P is problematic. “Gross” pretends that we can ignore destruction, including destruction of the environment, “Domestic” ignores the fact that we cannot consume what doesn’t belong to us, and “Production” should alert us to the fact that it is neither income or consumption. If we want material wellbeing, NNP would be better, and Actual Individual Consumption better still. A spectacular case comes from Ireland in 2015, whose GDP increased by 26.3 percent over the previous year, or 32.4 in current prices.1 This is not some error that arose from mismeasurement in the national accounts; there was nothing wrong with the numbers. Nor is it an indicator that the Irish were all of a sudden better-off. What happened was a large inflow of intangible assets by multinationals seeking Ireland’s generous tax advantages. Unlike physical plant and equipment, these intangible assets—software systems, patents, or brand names—can move instantaneously, at the stroke of a pen (or a keyboard.) They can also be removed instantaneously. With the rise of intangible assets relative to tangible assets, with the rise of international competition on tax breaks, and with the rise of globalization, these movements, and the changes in GDP associated with them, are becoming larger over time, and will swamp the more mundane changes in household disposable income or consumption that relate more directly to material wellbeing. If we insist on thinking of per capita GDP as a measure of material wellbeing, then the richest places in the world are soon going to be tax havens with small populations.

Again, there is nothing wrong with this situation, at least statistically, but if the news media and politicians continue to valorize GDP or per capita GDP, the concept will lose repute, undermining public confidence in the national accounts. Much better would be to focus on what is happening to people, through their levels of disposable income or consumption. The loss of meaning is compounded by the large fraction of GDP that is imputed—as much as a half in some countries—so that there is a disconnect between what people see around them, and what the statisticians measure, or at least do not measure, but report.

Of course, no one actually receives GDP, nor even household income, and the credibility of the accounts would be enhanced if it were better recognized that different people receive different amounts, and that narrowly based economic growth, favoring the top of the distribution, or favoring an educated elite, is very different from broad based growth, which is experienced by more people. As income inequality has widened, and has attracted more attention and more anger, the demands for distributionally disaggregated national accounts has become more urgent.

Piketty, Saez and Zucman2 (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.3

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. Even more egregious is the portion of GDP that comes from opioid manufacturers addicting and killing people for profit.

Government provides benefits, such as healthcare through Medicare, and because we have no output measure, the benefits are measured at cost, which is what PSZ do. But if healthcare providers are plundering the system by carrying out unnecessary but profitable procedures, or by selling drugs that are ultimately harmful, the costs are counted as benefits to the recipients of Medicare and Medicaid even though we know, albeit not precisely, that they are worth much less than their cost. We thereby understate inequality and poverty. Indeed, we could reduce poverty by inciting providers to carry out even more useless procedures whose cost is counted as income to beneficiaries, not just by PSZ, but by the Congressional Budget Office.


  2. Piketty, Thomas, Emmanuel Saez, and Gabriel Zucman, 2018, “Distributional national accounts: methods and estimates for the United States,” Quarterly Journal of Economics, 133(2), 553–609.
  3. Fixler, Dennis, David Johnson, Andrew Craig and Kevin Furlong, 2017, Review of Income and Wealth,