I’m reposting here an article that Erald Kolasi and I wrote for Mint Magazine. The article discusses the problems with ‘real’ GDP and possible alternatives for measuring economic activity. The published version is under a paywall, so I’m posting our (unabridged) manuscript here.
Why We Should Abandon Real GDP As A Measure of Economic Activity
Erald Kolasi and Blair Fix
For all that it purports to say, Gross Domestic Product (GDP) fails to explain anything relevant about the world. It values social services as it might domestic housework. It ignores ecological degradation by adding up all costs, regardless of what they’re for. It says nothing about the distribution of income and wealth. It also tells us little about quality of life factors. Cuba has a much smaller GDP per capita than the United States, but Cuba’s life expectancy is higher, or at least comparable, to that of the United States.
These issues are much discussed. But to understand why ‘Real’ GDP, in particular, reveals so little about the state of any economy, you need to dig deeper. In this article we expose the deep underlying technical flaws of GDP calculations by explaining some of the aggregation issues that plague nearly all macroeconomic theories. We also propose other metrics for thinking about the scale of economic activity.
The Meaning of Units
Growing up, you probably obsessed over your height and how fast it was changing. Every birthday, you stood tall against the wall, had a family member extend a tape measure, and listened impatiently for that special number. 3 feet and 7 inches. 4 feet and 2 inches. 5 feet and 4 inches. Growing up so fast!
Now imagine this. One day your devious brother decides to pull a trick by changing the tape measure used to measure your height. The new tape measure looks the same as the old one, except that the tick marks for inches are slightly farther apart. Your brother measures your height and reports that you haven’t grown at all. In fact, you’ve shrunk! You’re devastated, until you learn of your brother’s trick. You haven’t actually shrunk, you find out. It just appeared that way because your brother changed the unit used to measure your height.
The lesson here is that for measurement to be accurate, the units must be stable. When you measure your height, you want the measurement to change because you actually grew taller, not because you changed the length of an ‘inch’.
Using stable units is essential for objective measurement. Because of this, natural scientists have spent centuries refining their units. The kilogram, for instance, was recently redefined in terms of Planck’s constant, the ratio of the energy in a photon and its frequency that never will change. That replaced the block of platinum/iridium alloy that had been used to define the kilogram since 1889. The goal was to make this unit as immune to change as possible. This ensures that changes in the measured mass of an object reflect actual physical changes, not variations in the unit of measure.
When units are unstable, in contrast, objective measurement becomes impossible. When a unit varies, over time or space, we can’t be sure that measured variation reflects actual variation, or just variation in our unit of measure. This unstable-unit problem plagues many measures in economics, especially ‘real’ GDP, which is often reported with exquisite precision. Unlike natural scientists, however, economists are not in the business of carefully defining units using universal physical constants. Economists instead use prices, a social construct, as their unit of analysis. To measure economic growth, they add up the market value of all new commodities produced in a year.
The problem is that prices are unstable units of measurement. Relative prices between commodities vary wildly over time. This instability means that prices fail the only requirement of a good unit — to be uniform over time. This problem is acknowledged by government experts, yet remains largely hidden from the public in reports about GDP from the government and from the media at large. Instead of reporting the severe uncertainty in ‘real’ GDP, governments report a single official value. This value hides a myriad of subjective decisions that are used to ‘correct’ for unstable prices. Worse still, these decisions are strongly influenced by neoclassical economic theory, which describes a capitalist utopia that has no connection whatsoever to the real world.
Instead of wasting time with a useless quantity that reveals nothing profound about the world, we should seek new pluralistic methods for understanding aggregate economic activity.
The Problem of Unstable Units
The parable of your brother switching the tape measure when measuring your height illustrates the basic problem with unstable units. When units change over time, it wreaks havoc with objective measurement.
To get an idea of how this problem affects ‘real’ GDP, we’ll use another simple example. Suppose you’re a farmer who grows apples and oranges. Taking a hint from economists, you decide to measure the scale of your fruit production using market value.
In your first year of farming, you grow 10 apples and sell them for $1 each. You also grow 20 oranges and sell them for $2 each. The value of your total fruit production is $50 (10 apples × $1 + 20 oranges × $2). The next year you grow 5 apples and sell them for $1.5 each. You also grow 25 oranges and sell them for $1 each. The value of your total fruit production is $32.50 (5 apples × $1.5 + 25 oranges × $1).
Here is the question you want to answer: how much has your production changed? It’s tempting to answer that production has shrunk by 35% (from $50 to $32.5). But this simple answer neglects the fact that our units, prices, are unstable over time.
Faced with this instability, you take another hint from economists. You introduce a distinction between nominal and real prices. Nominal prices represent the market price, which changes with time. Real prices, in contrast, ‘correct’ for the effects of price change.
In our farming example, the nominal value of fruit output shrank by 35%. But what about the real value of output? To calculate the latter, we pick a ‘base year’ and project these prices into other years. We then measure output using these constant prices.
Let’s do this exercise using prices in the first year. We fix the price of apples at $1 and the price of oranges at $2. The real value of production in the first year remains unchanged from its nominal value ($50). But in the second year, the real value of production is $55 (5 apples × $1 + 25 oranges × $2). Using this base year, we find that ‘real’ fruit output grew by 10% (from $50 to $55).
Now suppose we use the second year as our base year. We fix the price of apples at $1.5 and the price of oranges at $1. The real value of production in the first year is now $35 (10 apples × $1.5 + 20 oranges × $1). The real value of production in the second year remains unchanged from its nominal value ($32.50). Using this base year, we find that fruit output shrank by 7% (from $35 to $32.50).
Do you see the problem? Depending on the year in which we fix prices, the ‘real’ value of fruit output either grows by 10% or shrinks by 7%.
How can this be? Our contradictory results are caused by unstable units. Relative prices vary over time. When we choose different base years, we change the weighting we assign to apples and oranges as we add them up.
If we use the first year as the base year, one apple counts as half an orange. Thus, the decline in apple output (from 10 to 5) is outweighed by the increase in orange output (from 20 to 25). But when we choose the second year as our base, one apple counts as 1.5 oranges. Now the decline in apple output outweighs the increase in orange output.
This is our changing meter stick in action. Our value for the growth of fruit production depends on the year in which we fix prices. The take-home message is that when relative prices change with time, the growth of ‘real’ value is ill-defined. A quantity that can be constructed to say anything really says nothing.
How Bad is the Unstable-Unit Problem?
The unstable-unit problem applies only if (relative) prices are unstable. That means we should ask — how unstable are prices? The answer, at least in the United States, is incredibly unstable.
Figure 1 shows price indexes for 10 commodities selected from the US consumer price index. Relative prices among these 10 commodities have varied wildly over the last 80 years. From 1935 to the present, the price of apples increased by nearly a factor of 50. Over the same period, the price of electricity increased by a factor of 7. And the price of televisions actually declined.
To show that we have not cherry-picked commodities, Figure 2 shows the price change of all commodities tracked by the US consumer price index. US prices, it seems, are incredibly unstable. To put this instability in perspective, the variation in relative price change is about 8 times greater than variation in male height.
This price instability translates into uncertainty in the growth of ‘real GDP’. While the US government reports only one official measure of ‘real’ GDP, it quietly maintains a database of ‘vintage’ GDP estimates. These are estimates calculated with different base years. Using this ‘vintage’ data, we can quantify the uncertainty in the growth of ‘real’ GDP caused by unstable prices. Figure 3 shows the results. Over the last 70 years, there is roughly 30% uncertainty in the growth of US GDP per capita.
Notice that the official measure of US ‘real’ GDP is at the upper end of the range of uncertainty. We doubt this is a coincidence. In fact, it is common for national governments to boost GDP growth by changing the base year. India recently showed a small increase in GDP growth by choosing a new base year. While this boost was small, it can sometimes be spectacularly large. Nigeria, for instance, recently changed its base year from 1990 to 2010. As a result, real GDP doubled, making Nigeria the largest economy in Africa. Base-year changes have led to similar boosts to GDP growth in Ghana, Kenya, Tanzania, Uganda and Zambia.
Almost everywhere you look around the world, GDP has become a lazy and reflexive way of justifying capitalist expansion and calling it “progress.” While the wealthy classes exploit struggling workers and hoard their riches in offshore tax havens, politicians are busy paying lip service to a useless number that reveals almost nothing about the world.
Picking Numbers Out of a Hat
When reporting ‘real’ GDP estimates, the behavior of governments is an interesting exercise in how not to do science. When there is uncertainty in a measurement, the appropriate response is to report it openly. Case in point, physicists are now engaged in an open debate about the expansion rate of the universe. Estimates based on ancient light coming from the Big Bang give an expansion rate that is different than estimates based on the movement of galaxies. This uncertainty is causing some physicists to question the standard model of cosmology. This is the right response to measurement uncertainty: reporting and debating conflicting results.
The wrong response to uncertainty is to pick a single value and declare it the ‘truth’. Yet this is exactly what governments do when reporting ‘real’ GDP. The instability of prices means that there is a large range for the possible growth of GDP. Yet governments do not report this range of uncertainty. Instead, they report a single measure of GDP that is based on an arbitrary choice of base year (or similarly arbitrary methods for ‘correcting’ for inflation). When GDP estimates are revised, the old estimates quietly go away.
Hiding the Problem
For the most part, economists are aware that there is a problem with their holy dogma. When speaking to fellow members of the clergy, they admit that ‘real’ GDP calculations are essentially arbitrary. In a 1995 paper, the economist Charles Steindel had this to say about calculating ‘real GDP’:
The economy consists of millions of individuals and firms producing a multitude of goods and services. This complexity virtually ensures that any method of estimating ‘real GDP’ involves making some more or less arbitrary decision about the most appropriate way to add up data from individual sectors.
Steindel hits the nail on the head. Methods for calculating ‘real’ GDP are arbitrary. Worse still, the arbitrary choice of method affects the growth of ‘real’ GDP. Again, this is common knowledge among government economists. In Chapter 4 of the NIPA Handbook, the US Bureau of Economic Analysis notes:
The fundamental problem confronting the efforts to adjust GDP and other aggregates for inflation is that there is not a single inflation number but rather a wide spectrum of goods and services with prices that are changing relative to one another over time. The index numbers for the individual components can be combined statistically to form an aggregate index, but the method of aggregation that is used affects the movements of the resulting index.
It seems, then, that government economists are aware of the problems we raise here. Arbitrary decisions about how to ‘correct’ for changing prices affect the resulting growth of ‘real’ GDP. Again, this stems from the fact that prices are an unstable unit of analysis.
Our response to this problem is to conclude that the growth of ‘real’ GDP is fundamentally uncertain. Or perhaps a better word is ambiguous. When scientists speak of ‘uncertainty’, they tacitly assume that there is a true value on which to converge. Two centuries ago, for instance, there was large uncertainty in the speed of light. But over time, better instruments allowed measurement to converge on a single value. This value, scientists believe, is the true speed of light.
But with ‘real’ GDP, there is no ‘truth’ on which to converge. Because ‘real’ GDP is based on an unstable unit (prices), its value is fundamentally ambiguous. Different methods of ‘correcting’ for inflation yield different values for the growth of ‘real’ GDP. And as Steindel admits, the choice of inflation-adjusting method is fundamentally arbitrary. The conclusion we reach is that the growth of ‘real’ GDP is inherently ambiguous.
Faced with the same evidence, however, mainstream economists reach a very different conclusion. Their response is to simultaneously admit that calculating ‘real’ GDP requires arbitrary choices, but then to report a single value as though it was the ‘truth’.
In the technical literature on inflation, for instance, there is an endless debate about the appropriate price index for quantifying inflation. A price index takes a basket of commodities and calculates the average rate of inflation among these items. But as
Francis Edgeworth noted more than a century ago, “There are as many kinds of average as there are purposes.” Thus, there are an endless variety of price indexes (the Paasche index, the Laspeyres index, the Fisher index, the Walsh index, and so on). Each index will give a different value for the rate of inflation, and hence, a different value for the growth of ‘real’ GDP. Again, this illustrates the ambiguity in ‘real’ GDP.
But when it comes to calculating the official measure of GDP, this ambiguity gets suppressed. Instead, government economists arbitrarily choose a single method. For instance, the US government currently calculates ‘real’ GDP by adjusting nominal GDP with an aggregate index formed through the multiplication of successive Fisher indexes in adjacent time periods. In popular parlance, this method is called ‘chain-weighting’. Rather than choose a single base year in which to fix prices, chain-weighting uses a technique that resembles a rolling base year. The method is meant to simulate the effect of changing prices and spending patterns over time.
This method was adopted in the mid-1990s. The official justification was that structural changes in the US economy, especially rapidly falling computer prices, compelled the government to end the fixed base year method. But contrary to the BEA’s assertions, this change did not produce a more “accurate” estimate. Such a statement implies that a true value of ‘real’ GDP exists independently of the techniques used to measure it, which is simply false, as we have shown above. Whether using a fixed or a rolling base year, ‘real’ GDP is still an arbitrary statistical construct. It is defined by the subjective choices of government economists.
A Regressive Measure of Progress
We live in a world where ‘real’ GDP is the dominant measure used to quantify economic progress. Mainstream debates focus on how to achieve economic progress, but the issue of how best to measure economic progress flies mostly under the radar. It’s simply assumed, by most sides of the political spectrum, that the growth of ‘real’ GDP is what constitutes economic progress.
We think this is a mistake. When socialists and the Left use ‘real’ GDP as a measure of progress, or even as a quantity of empirical analysis, they are implicitly buying into the very capitalist paradigm that they otherwise reject. Let us explain.
The use of ‘real’ GDP to measure economic progress is based on ideas from neoclassical economics. According to some of the dominant interpretations of neoclassical theory, the price of a commodity reveals the utility (i.e. pleasure) that a consumer derives from it. When we aggregate the value of all commodities, neoclassical theory posits that we are aggregating the utility of the entire society. When ‘real’ GDP goes up, in other words, so does aggregate utility. Everyone becomes better off!
Except they don’t.
We’ve already shown that the instability of prices means the growth of ‘real’ GDP is ill-defined. We simply can’t say for sure how much aggregate utility is increasing.
Second, we know that the growth of ‘real’ GDP correlates poorly with other measures of human well-being. Many ecological economists posit that this is because GDP treats everything with a price as contributing positively to society. Again, this comes down to the assumption that all prices reveal utility. If machine guns sell for the same price as MRI machines, neoclassical theory tells us that both contribute the same utility to society.
In response to such absurdities, ecological economists have developed alternative indicators that subtract the value of social ‘bads’ from the value of social ‘goods’. While well-motivated, this approach still assumes that we can aggregate the ‘real’ value of ‘goods’ and ‘bads’. But because prices are unstable, this aggregate is still ill-defined.
The central issue here is that the tenets of neoclassical economics are fundamentally flawed. To equate market value with aggregate utility, a host of assumptions are required. Consumers must be identical. Individual preferences must be independent of income and fixed over time. And markets must be perfectly competitive and in equilibrium.
When we equate ‘real’ GDP with social welfare, we are buying into a hopelessly flawed theory. But more than that, we are buying into market ideologies supportive of capitalism. Neoclassical economic theory has never resembled a scientific enterprise. It’s simply an ideology presented through an avalanche of mathematics. The underlying assumptions of neoclassical theory all serve to justify the capitalist status quo. When we equate market value with utility, we implicitly assume that individuals’ income indicates their contribution to society. The exorbitant pay of CEOs is seemingly fair. It apparently has nothing to do with their power within the corporate hierarchy.
‘Real’ GDP is a regressive measure of social progress. Not only is it ill-defined and based on flawed premises, it equates market value with social welfare. This justifies the income of the powerful. For the Left, this kind of worldview is unacceptable.
Alternatives to ‘Real’ GDP
‘Real’ GDP belongs in the dustbin of history. In its place we should adopt a plurality of measures.
We believe it is important to distinguish between economic distribution and economic scale. ‘Real’ GDP assumes that prices can be used to measure economic scale. In contrast, we assume that prices do nothing of the sort. Prices are a tool for distributing resources. The proper place for prices, then, is for understanding economic distribution.
The game of distribution is all about the command of market value. If a lawyer can charge more for his services than a janitor, the lawyer wins the game of distribution. This is not about productivity. It is largely about power. For instance, individuals’ income within firms correlates strongly with their hierarchical power — their control over subordinates. Political economists Jonathan Nitzan and Shimshon Bichler convincingly argue that the capitalization of firms indicates their ‘differential power,’ which is their ability to use property rights to further their own economic advantage over workers and other peer competitors. Dominant classes and corporations exploit this differential power to adjust wages, profits, and prices as they see fit, gaining more power over labor along the way.
When we treat prices as a tool for distribution, the proper thing to do is to compare prices. Nitzan and Bichler call this differential analysis. For instance, we can compare the nominal market value of different firms, or different groups of firms. Likewise, we can compare the income of individuals or groups of individuals. The meaning is in the comparison, not the aggregate value itself. The focus here is on how relative prices change over time, not what they reveal about the ‘real’ sphere of production (they reveal nothing).
But if prices are used solely for studying distribution, we must find an alternative dimension for studying economic scale. There are many possibilities. The choice of dimension should depend on our goals.
To measure the scale of the economy, we think it is appropriate to focus on energy. Physicist Eric Chaisson argues that energy is the universal currency of science. By measuring economic scale using energy, we put economics in line with the rest of science. And if we are concerned with sustainability, there is no better starting point than to focus on energy use. After all, the profligate use of fossil fuels under a capitalist economy is the primary driver of climate change.
Energy has many forms as it is flows through society. One possibility is to focus on primary energy consumption, and see how this relates to changes in social structure. Another possibility is to measure ‘useful work’ — the consumption of end-use energy. Still another possibility is to measure the aggregate flow rate, which is a measure of all annual energy conversions in an economic system.
The key, we believe, is to separate the study of economic distribution from the study of economic scale. The former is the appropriate domain of prices. The latter is best measured using biophysical units. We do not mean to suggest that scale and distribution are causally unrelated. We are simply highlighting the point that any connection between the two is vastly more complex than what has been traditionally recognized in economics.
Science Is Only as Good as Its Measurements
For much of the 18th century, scientists explained combustion in terms of an invisible substance called phlogiston. Combustible materials were thought to contain phlogiston, which was released when burned. The phlogiston theory was the accepted explanation of combustion from the 1660s to the 1770s. The theory was discredited only when Antoine-Laurent de Lavoisier accurately measured the gases of combustion, showing that they had mass. The lesson of phlogiston theory is that our understanding of the world is limited by our ability to accurately measure it.
Echoing Paul Romer, we believe economics is still stuck in a phlogiston state. ‘Real’ GDP, the primary metric used to study macroeconomics, is hopelessly flawed. It supposes that prices, a social construct, can be used to objectively measure the growth of ‘real’ economic output. The problem is that prices are an unstable unit, a changing meter stick. This means that the growth of ‘real’ GDP is fundamentally ambiguous. Worse still, the very concept of ‘real’ GDP as a measure of progress is based on the vapid and idealized assumptions of neoclassical theory, repeatedly shown to be false.
We advocate separating the study of distribution, which should focus on prices, from the study of economic growth, which should focus on biophysical flows. This new approach will push economics into a post-phlogiston state. This stance has important political implications as well, especially for those who seek to challenge the reigning status quo. Continuing to use ‘real’ GDP as a measure of social progress implicitly accepts a theory (neoclassical economics) that has long been used as an ideological justification for capitalist power.
The time has come to discard ‘real’ GDP and to elevate new measures that actually address important social problems.