Some Problems With GDP

The price of everything and the value of nothing

Many people use GDP to make claims about changes in people’s welfare over time. This is entirely reasonable, and would likely be correlated with any reasonable measure of people’s welfare. It is not, however, perfect, and it will miss things. First, it often does not include many things which we might find valuable, such as leisure, life expectancy, total population, and household production, and may not track onto utility if there is inequality and non-linear utility as a function of wealth. Second, GDP cannot adequately capture the gains to utility from free goods, in particular the Internet. Relatedly, goods with markups will not be able to state. Third, the way in which we treat investment arbitrarily increases GDP in places with higher capital intensities. The first two are actually quite intimately related to each other, as I will show.

GDP is intended to be the resources available for consumption. It does not take into account our productivity in making it. If a country works longer hours than another, they will have a higher GDP, but we cannot say that they are really better off. Likewise, if we live longer we will have more time in which to consume, but we would not pick up on this except insofar as it affects our total national consumption. Jones and Klenow (2016) is one such example, where they ask how much you would have to compensate someone to accept living in a country which has all the traits of another, but for, say, life expectancy. You can also do something similar for positive and negative externalities which do not turn up in changes in production. They arrive at this by fitting equations representing a utility function using microdata. Later work by them and others considers different ethical beliefs around having more people alive. The same GDP per capita, but with more people around to experience it, is probably better, and so Adhami, Bils, Jones, and Klenow (2025) argue that the growth in poor countries in underrated. I don’t necessarily believe the value that we put on a human life, but it’s a fine starting point. Ultimately what we need to do is put implicit prices on things which don’t have explicit ones.

We do this for some stuff. Household production is estimated by the Bureau of Economic Analysis, in parallel to GDP, in the household production satellite account (or HPSA). This is fairly simple compared to the other things we’ll have to estimate later, as much of what we do in the house has an implicit price in the market which we can compare to. Providing childcare can be done by a parent or by a nanny, and cooking can be done by the family or by a restaurant, all of which have prices. (As far as I know, we do not impute the benefits of lovemaking in the household by consulting the price of prostitutes, although I understand the staff of the BEA to be quite keen on the idea). This can be easily integrated into national accounts, and especially when proportions shift as during Covid, will greatly shift your understanding of what happened to actual consumption. This also leads to the apparent advantage of Europe in leisure being somewhat illusory. They do more work at home, and their leisure isn’t really leisurely.

For stuff with no obvious market price, a reasonable approach is to argue, for some goods, that we don’t really need to. Suppose Wikipedia replaced our spending on encyclopedias. Holding working hours fixed, we would simply spend more on other items, which will show up in GDP. If we reduce our working hours, then this returns to leisure as before, and we can apply the same work. What we will miss, though, is any excess utility from the free service. We can’t measure the consumer surplus.

Related to this are the effects of markups on welfare. Imagine a monopolist selling a single good. They will curtail production to maximize the profit they get. We can easily imagine a world in which the monopolist faces a marginal cost curve such that total expenditures at the profit maximizing price is the same as one where price is at the marginal cost. Since we find GDP from expenditures, we would say that these two scenarios have identical GDP. The internet is simply a special case of this, where the marginal cost is approximately zero, and thus any distortion away from marginal price (which would be a reduction in welfare) shows up as an increase in GDP.

One method of including them is to try and look at the time spent using a service. Goolsbee and Klenow (2006) is an early paper along this line. They argue that, at the margin, people allocate their time in such a way as to maximize utility. If people choose to consume the internet instead of doing other things, they must regard it as better. Unfortunately their data is not necessarily the best – I do not believe that people are trading off only between working and using the internet – but the method allows us to put numbers on how much people value the internet (which they estimate, in 2006, to be about $3000 per person).

This is similar to the argument of Feldstein (1999) about finding the deadweight loss from income taxation. There are a few ways in which we can change our behavior to avoid income taxes – we can reduce our hours worked, obviously, but we can also change the form of consumption. Since these must have the same utility at the margin, all we need is the total change in taxable income.

I believe a similar approach can be used for leisure. If countries work different amounts due to differences in tax rates, and not due to differences in preferences, then we can perhaps abstract away from leisure entirely. Of course, this would seem to imply that the preference of consumers is to work all hours, and we don’t do so only due to taxes. In any event, this has been a detour to be picked up later – back to sounder footing.

You can also conduct experiments to discover the value which people place upon a good. Brynjolfsson, Collis, Diewert, Eggers, and Fox (2019) simply ask people what they are willing to pay for things such as the camera in a smartphone, or how much they would need to be paid to give up a good. Note again that it is ambiguous as to whether this over or underestimates the benefit to society, as these free goods – or really any goods – can have positive or negative externalities, but you at least capture the value to the user. They estimate that Facebook alone has added between .05 and .11 percentage points in growth annually to the United States.

The last objection to GDP is its treatment of investment. GDP counts only final goods, not intermediates. We would very obviously not want to arbitrarily inflate GDP by counting the value of iron ore and pig iron and steel and railroad tracks separately. However, many final goods are in fact intermediaries to the production of something else later. Take the railroad tracks – they will enter GDP once when we take their value as a finished good, and then again through boosting production by transporting goods hither and thither. A country which has relatively more investment will have arbitrarily higher GDP than one which has less.

This problem is brought up by Robert Barro in “Double-counting Investment” (2020), which is a generalization of the argument of Koh, Santaeulalia-Llopis, and Zheng (2019) that the observed decline in the labor share in income is actually entirely due to including intellectual property in national income statistics, instead of as an intermediate good. Capital isn’t actually getting all that much more income – instead, we’re counting the payments to produce these intermediate goods as final goods.

We already adjust for this with housing. We don’t count houses when they’re produced toward GDP. Rather, we view the house as an investment which produces a flow of lodging. We estimate what it would cost to rent the place, and have that be the contribution of the house to GDP for each period.

The main difficulty is that it is really, really hard to separate out consumption from investment. Suppose someone buys a truck. Part of the value of a truck is that you can ship goods from one place to another, which is an intermediate input into other goods. We wouldn’t want to count goods which are shipped around a lot as more valuable than goods which aren’t shipped around, so we deduct it out of GDP; but at the same time, we also buy trucks because they’re cool and honk-honk and vroom-vroom and wheeee and big truck makes me big man. We don’t know how much of our purchase was due to each part, and thus cannot subtract out only part.

This has been a discussion of how GDP can be improved. I would caution, however, against shallow cynicism that denies any usefulness of GDP. Any reasonable measure of economic welfare will be correlated with GDP. As Jones and Klenow note, their adjusted measure has a correlation of 0.98 with GDP. The rich countries of the world are indeed much richer than the poor countries, and the poor countries have really grown by a ton. Including other sources of gains into GDP tends to increase, not decrease, the gains which poor countries have experienced, and shows them as being closer to the West than GDP would indicate. For example, while GDP has not consistently improved in Africa, life expectancy has greatly increased. GDP isn’t perfect, but that’s hardly a reason to think the world hasn’t gotten better.