The History of GDP

In 1931, the United States faced a strange void. The Great Depression was entering its third year, yet the nation's most recent national income estimates only went up to 1929. Everyone could see factories shutting down and workers losing their jobs, but there were no numbers to tell just how much had been lost.

In June of the following year, 1932, Senator La Follette of Wisconsin introduced Senate Resolution 220. It directed the Department of Commerce to report the national income for 1929, 1930, and 1931. The assignment landed in the Division of Economic Research within the Bureau of Foreign and Domestic Commerce, and Simon Kuznets, a researcher from the National Bureau of Economic Research serving as an advisor, took charge of the entire project.

The estimates were completed in late 1933. A 261-page report titled *National Income, 1929–1932* was submitted to the Senate in January 1934 and published as a Senate document. Then something curious happened: this dry statistical report became a bestseller. U.S. Department of Commerce records note that over the eight months following its publication, its sales surpassed the Statistical Abstract, which had long been the government's flagship publication.

The structure of the report is also noteworthy. The opening chapter explains what is counted and what is left out, as well as how these figures must not be interpreted. Tables by industry and type of income follow, and the appendix details exactly which sources each figure came from and how it was calculated. This practice of fully disclosing every source became the benchmark for all future estimates.

Starting in February of that year, the Commerce Department's monthly journal began publishing these estimates annually. What began as a one-time request to tally the numbers had turned into a permanent operation. Kuznets himself later moved on to the other side, becoming a user of these very numbers. At the War Production Board, he relied on these estimates to assess whether wartime production targets were actually feasible. This was the starting point of the figure we now see in the news every quarter.

There isn't just one answer to this question. Depending on what kind of "first" you are looking for, the person, the country, and the year all change. Once you divide the story along five axes, each piece settles neatly into its own place without contradiction.

If you mean "measuring an entire nation's income by congressional mandate and submitting it to the government," that was the United States in 1934, and the creator was Kuznets. If you mean "building national accounts using double-entry bookkeeping," that was Britain in 1941. In a white paper on war finance, Richard Stone and James Meade attached an appendix on national income and expenditure; dividing the economy into four sectors—households, businesses, government, and the overseas sector—and setting it up so that one sector's spending showed up as another's income marked a decisive turning point. Because the pieces interlocked, figures could finally be cross-verified.

If you mean "becoming an international standard," that was the first edition of the United Nations System of National Accounts in 1953. Stone served as chair of the expert group, built on a 1947 League of Nations subcommittee report and the 1952 standard system from the Organisation for European Economic Co-operation. If you mean "the standard's first major overhaul," that was 1968.

One thing stands out here. Both the United States in 1934 and Britain in 1941 built these systems because they were facing urgent crises. One needed to know how badly things had collapsed, while the other needed to know how much war funding could be extracted and from where. Stone subsequently served as chair of the League of Nations subcommittee, director of the Organisation for European Economic Co-operation unit, and chair of the UN expert group, turning methods born of emergency into an international standard.

The fifth axis is the most important in this story. It was in 1991 that the United States switched its headline indicator from Gross National Product to Gross Domestic Product. Gross National Product had held the flagship role since 1934, but international standards gave priority to Gross Domestic Product, and because other nations were already using it, international comparisons were getting distorted. The former counts what a nation's people earn wherever they are; the latter counts whatever is earned inside the country, whoever earns it. They are two different rulers measuring different things.

There is a widespread story that after Kuznets created GDP, he warned people not to trust it. If you open the original document, that is half true and half false. That passage really does appear in the opening pages of his 1934 report. The section heading is "Uses and Abuses of National Income Measurements."

Yet the section does not begin with a warning. First, he writes this: adjusted for price fluctuations and divided by population, national income reflects the movements of a country's economic welfare. Then he adds qualifications—on two levels. First are the caveats when reading this figure as an index of productivity: because this estimate primarily counts activities with a market price, things like unpaid domestic labor are left out; market prices vary depending on the distribution of income in that country; and the method of distinguishing between a company's gross and net income follows the accounting conventions of the time. Then he writes that difficulties multiply when it is read as an index of welfare. Without knowing how income is distributed, one cannot properly measure welfare, and no estimate measures the reverse side of income—the toll and effort of labor. Only then does the sentence come: the welfare of a nation can scarcely be inferred from a measure of national income such as that defined above.

One of the most widely quoted statements is not in that report at all. It is the passage suggesting that we must distinguish between quantity and quality of growth, and specify growth of what and for what. When you download the full text of the report and search through it, that phrasing does not appear a single time. It is said to have come from another piece he wrote for a magazine decades later, though we could not verify that original text here. So all we can say is: "It is not in that report."

What is even more interesting is what he left out. Domestic labor, the services derived from consumer durables owned by households, odd-job earnings, relief and charity, asset price fluctuations, and income from illegal activities like theft, drug trafficking, and bootlegging. His reason for excluding the last category was that socially harmful activities cannot be considered services. He noted himself that some items were excluded simply because data was lacking, and added the caveat that owner-occupied housing might be an exception to his reasoning. This list was not a fixed law of nature, but a judgment of its time. As we will see later, the list was revised.

In fact, one item on that list has already been overturned. The practice of assuming homeowners pay rent to themselves and counting that value as production is called imputed rent, and today it is the single largest imputed item in the U.S. national accounts. From 1996 to 2006, it accounted for 6.0 to 6.2% of GDP. The reason is that it is necessary if we want to measure rented homes and owner-occupied homes with the same yardstick. The part about illegal activities was also reversed, but we will see that in the next scene.

There is an old fork in the road in accounting. If you view something as used up within the year, it is treated as an intermediate input and is not added separately to final output. If you view it as buying something used over multiple years, it counts as investment and enters GDP. Research and development spending sat right on this line.

On July 31, 2013, the U.S. Bureau of Economic Analysis moved that line when it comprehensively revised the National Income and Product Accounts. It began treating money spent on research and development by businesses, government, and non-profit institutions as the purchase of fixed assets. Entertainment, literary, and artistic originals also became assets, and a new asset category called "intellectual property products" was created. Since there was no market price from buying and selling them, their value was calculated by summing up the costs to produce them.

The impact was massive. The 2012 current-dollar GDP was revised upward by $559.8 billion, and $526.0 billion of that was due entirely to changing the definition. Capitalizing research and development alone accounted for $396.7 billion, and entertainment, literary, and artistic originals accounted for $74.3 billion. The remainder came from capitalizing ownership transfer costs for residential assets at $42.3 billion, and treating defined-benefit pensions on an accrual basis at $12.6 billion.

Other agencies expressed this same overhaul differently. Eurostat recorded in its own tables that the level of U.S. GDP in 2012 had risen by 3.6%. The U.S. Bureau of Economic Analysis showed it as a dollar amount, while Eurostat presented it as a percentage. Because two agencies recorded the same event in two different ways, we shouldn't blend the two together. This kind of caution will continue to be necessary for the numbers ahead.

Not a single new thing had actually been produced that year. The research labs were the same, and the movies were the same. What changed was simply which box on the ledger those activities were entered into. This is where "the country grew bigger" and "the country produced more" parted ways for the first time.

Europe implemented the new national accounts standard starting in September 2014. It was the European version of the international standard adopted by the UN in 2008, sharing the exact same roots as what the United States had done in 2013. The rule counting research and development expenditures as investment was included here as well, and spending on weapons systems also became investment.

On October 17 of that year, the European statistical office published the results by country. Based on 2010, the GDP level of the EU28 rose by 3.7%, of which 2.3%p came from methodological changes and 1.4%p came from introducing new data. Within the methodological changes, research and development accounted for 1.9%p and weapons systems made up 0.2%p.

The interesting part was that the magnitude varied by country. The country whose 2010 level saw the largest increase was Cyprus (9.5%), followed by the Netherlands at 7.6%. Luxembourg was limited to just 0.2%, and Latvia actually went down by 0.1%. Looking at methodological changes alone, Sweden (4.4%) and Finland (4.2%) saw the biggest gains—both driven mostly by research and development—while Bulgaria stood at 0.4%. Countries that invest heavily in research grew noticeably larger under the new rules. In other words, when you change the ruler, the distance between countries changes too.

Yet the growth rates stayed virtually untouched. From 1997 to 2013, the change in the annual growth rate hovered around 0.1%p. The level shifted dramatically while the speed remained unchanged—a contrast that will keep resurfacing throughout this story.

There was one more thing. Several countries took this opportunity to overhaul how illegal activities were recorded in the accounts. Eurostat's explanation went like this: the line determining what counts as production is not determined by whether it is legal or not. So in principle, drugs, smuggling, and prostitution are included. However, the same document added that some countries could not actually include them due to a lack of data. It was the very same category that Kuznets had excluded eighty years earlier on the grounds that it was socially harmful.

This is the heart of the story. Across the world, there are so many rules that use GDP as their denominator.

The European Union's fiscal rules set reference values of a 3% deficit and a 60% government debt ratio against GDP at market prices. That is written into Article 1 of Protocol No. 12 annexed to the Treaty on the Functioning of the European Union. At the 2014 Wales Summit, NATO reaffirmed the guideline to spend at least 2% of GDP on defense, and at the 2025 The Hague Summit, agreed to raise it to 5% by 2035—with at least 3.5% for core defense requirements and up to 1.5% for broader defense and security investments. The international aid target was set at 0.7% of gross national product by a UN General Assembly resolution on October 24, 1970. All four use the size of a country as their denominator.

That is why this happened in the fall of 2014. On October 21 of that year, the European statistical office published fiscal figures under the new standards for the first time, and the euro area's 2013 government debt ratio had fallen from the 92.6% reported that spring to 90.9%. For the EU28, it dropped from 87.1% to 85.4%, and the fiscal deficit narrowed from 3.3% to 3.2%. The report card for a year that had already passed was retroactively improved.

The European Central Bank broke down this decline in a table, and the result was surprising. The new standard itself actually raised the ratio by 0.9%p. That was because more institutions were classified into the government sector. Other revisions also added 0.3%p. However, the effect of an expanded GDP level pulled it down by 2.8%p, so overall, the ratio dropped. In Ireland, Cyprus, and Slovenia, the increase in the GDP level alone improved the fiscal balance ratio by more than 0.4%p.

The same thing happened with the aid target. In 1974, Sweden became the first country in the world to meet the 0.7% target, but when its gross national product figures were revised, the milestone year was pushed back to 1975. The money the country spent had not changed by a single penny. By the way, the name of the denominator for that target changed once too. The 1993 System of National Accounts eliminated the term "Gross National Product" and replaced it with "Gross National Income."

There is an observation that when a measure becomes a target, it ceases to be a good measure. We call this Goodhart's Law. But we should read carefully here. There is no evidence in any primary source that statistical agencies altered the ruler to pass these rules. Europe's overhaul was an implementation of international standards, and Korea's benchmark year revision is routine work carried out every five years. Only one fact remains: because the ruler and the rules are tied together, whenever the ruler moves, the ratio used in the rules moves with it.

A base year sets the benchmark point in time for prices and industrial structure. If you leave it untouched for too long, industries that emerged in the meantime get left out, and their shares drift away from reality. That is why the UN Statistical Commission recommends updating it every five years. Nigeria had not updated it once since 1990.

In 2014, it finally did. It moved the base year from 1990 to 2010, adopted the UN's 2008 System of National Accounts and a new industrial classification, and compiled a supply and use table for the first time in the country's history. The International Monetary Fund noted that until then, constant-price series had relied on the consumer price index. It also expanded the pool of surveyed businesses by about ten times.

The numbers jumped dramatically. The level rose by 60.7% for 2010, 68.4% for 2011, 76.9% for 2012, and 88.9% for 2013. With its 2013 GDP recalculated at $509.1 billion, Nigeria became the largest economy in Africa and the 26th largest in the world. The denominator effect followed right away. National debt now looked like 10.6% of GDP, and the fiscal deficit appeared to hover around 1%. Yet the effect cut both ways—for the exact same reason, non-oil tax revenue now looked lower, dropping below 4% of GDP.

Many people back then claimed the numbers had been "inflated," but looking at the primary data tells a different story. In a document released in December of that year, Nigeria's National Bureau of Statistics itself noted that the rebasing actually resulted in lower sectoral and real growth rates, alongside a warning that simple comparisons with the old figures would cause misunderstandings. The World Bank went a step further. After the revision, annual agricultural growth fell to an average of 4.2%, which the Bank found aligned much better with the stubborn reality of rural poverty. It meant the previous estimate of nearly 6% had been out of step with reality.

There were reasons for having put it off for so long. In the same document, the statistics bureau noted that updating the base year requires large-scale surveys and supply and use tables, and recalculating time series prior to the base year is so difficult that few countries do it. For comparison, South Korea has performed this same exercise 13 times, once every five years. Each time, the increase in GDP varied—from 8.4% for the 1980 base year to 3.9% for the 1990 base year and 10.9% for the 2000 base year—averaging 7.3% up through the 2015 base year. The problem was not updating the numbers, but waiting so long to do it. Nigeria updated its base year again to 2019 in July 2025.

On July 12, 2016, Ireland's Central Statistics Office published the 2015 National Accounts. Constant-price GDP had jumped from €193.16 billion to €243.914 billion. That was 26.3% in a single year. The statistics office itself noted that an expansion of this magnitude was unprecedented in the history of OECD member countries.

The very next day, the same agency issued another press release. The gist was this: like all other statistical offices, we must publish GDP according to international rules. Yet because our economy is so open, this figure does not always help people understand what is actually taking place in Ireland. Please, look at other indicators as well. Alongside that message, they placed other figures on the same page. That year, household consumption grew by 4.5%, average annual employment rose by 2.6%, and retail sales volume increased by 8.2%. 26.3% and 4.5% were figures from the very same country in the exact same year.

What happened was that corporations had relocated assets, including intellectual property, to Ireland, which meant the income earned by those assets began to be counted inside Ireland. It was the direct result of measuring strictly by the book. And the denominator effect followed immediately. General government debt as a percentage of GDP fell from 105.2% at the end of 2014 to 78.6% at the end of 2015. The statistics office noted that this improvement was primarily driven by the expansion of that year's GDP, along with repayments of International Monetary Fund borrowings.

What the statistics office did was not alter the numbers. In September 2016, it convened the Economic Statistics Review Group. Policymakers, academics, labor, business, and regulatory bodies gathered, joined by Eurostat and the International Monetary Fund. Following its recommendations, a complementary metric called Modified Gross National Income was released for the first time in July 2017. It takes Gross National Income and subtracts the factor income of redomiciled firms, depreciation on intellectual property, and depreciation on aircraft owned by aircraft leasing companies.

That is why Ireland now has three figures standing side by side. In preliminary figures for 2018, Gross Domestic Product was €324 billion, Gross National Income was €254.2 billion, and Modified Gross National Income was €197.5 billion. It is not because any of them is fake. It is because the three measure different things. And the newly created metric does not replace Gross Domestic Product. It stands right beside it.

By this point, a clear rule emerges. Every GDP figure must come with: which agency published it, when it was released, and which version it is. Without these three details, you cannot compare it side by side with another figure. This type of information is called metadata, and on this ruler, it is not mere decoration—it is part of the value itself.

The statistical office of the European Union even compiled its own list answering "why the same figure differs across publications." Differences in revision dates, changes in the EU's member country composition between then and now, different currency units, whether seasonal adjustment is applied, different base years for chained volume measures, and whether the measure is in nominal terms or volumes. If even one of these six does not match, the value for the exact same country in the exact same year will differ. That very document also includes this note: values in paper or PDF reports are rarely updated after publication, while online databases are more up to date.

Rankings also shift depending on the ruler. To compare countries, you must align different currencies into one, and there are two ways to do that. Converting at market exchange rates makes countries with higher price levels look larger. Measuring by purchasing power parity (PPP)—which strips away price differences to compare real volumes—narrows that gap. According to the reference year 2021 results released on May 30, 2024, by the World Bank's International Comparison Program, world GDP measured in purchasing power was 152 trillion dollars, with China accounting for 18.9%, the United States 15.5%, and India 7.2%. If converted at market exchange rates, the order is not the same. This is why you must never blend both values in a single sentence.

South Korea's 2024 revision faithfully followed this rule. When the base year was updated, 2023 nominal GDP rose from 2,236 trillion won to 2,401 trillion won, and gross national income (GNI) per capita rose from 33,745 dollars to 36,194 dollars. The point at which GNI per capita surpassed 30,000 dollars was also moved up three years, from 2017 to 2014. While the report noted that the country's global ranking rose from 14th to 12th, it added a caveat right next to it: "provided that statistical revisions in other countries are absent or minor." After all, a ranking is not something decided by your movement alone.

Let me add one more thing. Determining an economic recession is also an issue of how you count. The rule of thumb that "two consecutive quarters of declining real GDP equal a recession" is widely used, but the committee in the United States that designates the peaks and troughs of the business cycle does not use that as its standard. The committee clarifies that while most recessions it designates follow that pattern, not all do. In practice, it evaluates monthly indicators together, such as real personal income, employment, and industrial production. On top of that, its official announcements come long after the fact because it waits for data revisions.

In March 2025, the 56th session of the United Nations Statistical Commission unanimously recommended the adoption of a new System of National Accounts. It marks the first major overhaul since the 2008 edition, centered on globalization, digitalization, well-being, and sustainability.

The biggest change is this: they decided to treat data as a produced asset. Marketing assets were included as well. They also updated how the ownership and depletion of natural resources should be recorded in accounting. How to calculate this in practice is still being developed. A joint task force formed by Eurostat and the International Monetary Fund, alongside an expert group from the Organisation for Economic Co-operation and Development, is preparing the guidelines.

In a way, data is about to go through what research and development went through in 2013. When the box where something is entered on the ledger changes, the size of a country changes—and when that size changes, the rules that use it as a denominator shift along with it. Yet because some countries amass massive amounts of data while others do not, we might see a rerun of what happened in Europe in 2014, where research-heavy economies suddenly grew larger. What actually happens will depend on how much and how each country implements the standard.

There are four things to watch. First, which countries implement it, and when. While the standard is set internationally, implementation timelines differ across countries, meaning that in the interim, we will be comparing figures measured by different yardsticks. Second, what happens to the rules that use GDP as a denominator. Keeping the thresholds unchanged makes passing easier, but modifying them triggers debates of its own. Third, whether more countries will create separate supplementary indicators, just as Ireland did. Fourth, whether the rebasing cycle for benchmark years is maintained.

We will not decide here which indicator is the right one. The debate surrounding alternative metrics is still ongoing, and what this story has shown is not the answer to that debate, but the fact that redefining the yardstick has been repeated time and again. That was ultimately what Kuznets meant when he wrote on the very first page in 1934: if you want to use this number, you must first understand what it was designed to count in the first place.