On February 14 2018, the Singapore Department of Statistics published its latest data on the country’s Gross Domestic Product (GDP). According to the press release, `for the whole of 2017, the Singapore economy grew by 3.6 percent’. Yearly GDP growth, as well as GDP per capita are perhaps the two most widely reported, cited, analyzed, and forecasted measures of national economies, and they often heavily impact the outcomes of elections, the direction of international investment flows, the extent of international aid a country is entitled to receive, and so forth.
There are obvious advantages to summarizing the state of a complex system like `the economy’ in a single number. One such advantage is that it enables countries to be ranked linearly. For instance, based on the 2017 ranking by the IMF, newsreaders and analysts could assert that ‘Singapore (3rd) is richer than Switzerland (9th), but not as rich as Qatar (1st)’. Another important advantage is that a single number seems to set a clear goal for elected politicians: `if you want to stay in power, maximize the nation’s GDP!’. However, Goodhart’s law – when a measure of a system is turned into a policy target, it ceases to be a good measure – applies here. Perhaps no other economic measure has been so intensely targeted for the purposes of control as the GDP. Here, we would like take a look (1) at the history of GDP calculation, and (2) at some surprising and questionable aspects of how GDP numbers are put together.
1. A cursory history of the GDP
One might be surprised by how recent national accounting is. While the invention of the idea of calculating national income can be traced back to the 17th century, due to the lack of reliable data, statistical tools, and institutional organization, we have no running calculations of national income in the modern sense until the 1930s. The original impetus for an aggregate statistic describing the state of the economy was provided by the Great Depression (1929-1941). Governments wanted to know the extent to which citizens became impoverished, and needed policy tools to improve the situation.
Figure 1: A graph compiled from Angus Maddison’s data comparing the GDP per capita of a few major economies since 1700 CE. [Figure by M Tracy Hunter.]
In the United States, the economist Simon Kuznets aimed at measuring economic welfare, rather than output. He argued that national income estimates were not reliable measures of economic welfare. For example, expenses incurred for military armament, advertising, or financial speculation should be deducted first. According to Kuznets, even expensive city housing, or the construction of subway systems should be excluded, since they also represented merely “an evil necessary in order to be able to make a living”. Kuznets’ ultimate goal was thus to devise a way to measure individual welfare, a concept that he dubbed ‘income enjoyed’, a hedonic measure which included even the subjective enjoyment of exercising one’s own productive powers.
Albeit Simon Kuznets later won the Nobel Memorial Prize in Economic Sciences, his methodology had less direct influence on policy than the work of Colin Clark in the United Kingdom. Clark was the first one to define national income as the sum total of all produced goods and services, weighted by their market price. Moreover, he introduced the idea of triple accounting for national accounts: Based on his work, the GDP of a country is calculated from the production, income, or expenditure side.
To calculate income on the production side, we need to add up the market value of all goods and services produced in a certain year, subtracting the market value of intermediaries, that is, anything used in the production of others goods/services – omitting the latter step would lead to double counting. On the income approach, the GDP of a country is composed by the remuneration of the factors of production: wages (for labour), interest (for capital), rent (for land), and residual profits. Finally, from the perspective of expenditures, the GDP of a country is divided between private consumption, government spending, investment, and net exports (exports minus imports). The three calculations have to lead to the same result, by definition (see Figure 2).
Figure 2: The circular flow and GDP calculation. Calculating GDP with the production approach means following the upper black arrow. The expenditure approach follows the upper red arrow. The income approach follows the lower red arrow. Note that the government is not represented. [Picture by Irconomics]
National accounting moved to the forefront of economic planners’ interests with the advent of the Second World War. For both Kuznets and Clark, government expenditure did not contribute to the GDP. It was John Maynard Keynes who would change this aspect of national accounting. For Keynes, government spending was a principal policy tool, as it enabled governments to intentionally increase economic activity, raise employment, restore market confidence, and thus put the economy on track in times of crisis. In his 1941 book, ‘How to pay for the war’, he argued that government expenditure should be included as a component of national income. Naturally, this was appealing to government officials: If government expenditure for armaments would have to be subtracted from national income, as per Kuznets and Clark, national income would have to fall during the war; if it can be included, national income could rise, despite possibly falling living standards. Indeed, both U.S. and U.K. GDP per capita rose during the years of their respective involvement in the war (by 43% in the U.S., and 13% in the U.K.).
After WWII, supported by statisticians and economists in the U.K., U.S., and the nascent United Nations, GDP quickly became the international standard for comparison of economic development. Between 1945 and 1970, world GDP has almost tripled. However, and partly due to this development, the meaning of the indicator itself became much more problematic, and the discrepancy between measuring economic welfare and GDP continued to grow.
2. The difficulties of calculating `the’ GDP
Note that the GDP is a gross number, because it doesn’t include the loss of value of the existing capital stock, like machines, infrastructure, buildings, etc. This is why it is possible for a GDP to actually rise during times of war or major natural disasters.
A related issue, prominent for environmentalists especially since the highly influential Limits to Growth report of 1972, is that the GDP does not include estimates for the decrease of stocks of raw materials, or natural degradation. According to this point, when a country consumes its stock of non-renewables, or pollutes its natural environment, it damages its future capacity to grow. Such a country is then not growing sustainably, and therefore a proper indicator for its growth should be lower than its GDP.
Clark already noted that as economies grow, their service sector contributes an ever larger share to the national economy. Now, the output of the agricultural and industrial sectors is relatively straightforward to measure: multiply the produced amount of corn/coal with the market price of corn/coal. However, it is completely unclear how we should measure the output of, say, an urban planner working in a government office, a kindergarten instructor, or a researcher of machine learning. We could estimate the value of their output through their respective wages, but that would mean that, by definition, the productivity of workers in service industries would always be 1, and could not rise. Or, we could develop completely arbitrary performance measures (say, number of reports emailed to a supervisor per month, or nursery rhymes taught times quantity of children), which will probably lead to ignoring the quality of the work, and terrible policies – recall Goodhart’s law. A specific methodological problem arises with the contribution of the financial sector, whose contribution would be estimated to be negative by a straightforward approach.
Everyone knows that not all work is remunerated by wages. There is scarce data on these activities, and they are therefore excluded from GDP calculations, leading to the oft-cited ‘paradox’ that the singleton who marries his/her housekeeper will reduce the GDP. Estimates show that various kinds of domestic work, taking care of relatives in need etc., valued at current market prices could more than double the conventional GDP figure. To this, we should also add non-paid community work, such as editing Wikipedia.
Industries or services which operate illegally (or semi-legally) will obviously not report (correct) numbers to a central statistical agency. To correct for this error, in many countries, the GDP includes estimates for the size of the shadow economy (the total value of drug production and trafficking, the value of prostitution services, corruption, and so forth).
Finally, the contribution of technological change to human well-being might be significantly underestimated by GDP numbers, because it simply values products at their market prices. For instance, antibiotics were initially extremely expensive to produce, but advances in chemical engineering have led to a sharp drop in its price. Using the quantity-times-price approach, the contribution of antibiotics to the national economy might drop as they become ubiquitous and very cheap to produce. Similarly, the price of personal computers may have been close to constant over the last thirty years, but their computational power has grown exponentially. The contribution of the Internet to our well-being might be severely understated by looking at how much we pay to our Internet Service Provider.
All these shortcomings of the GDP have led to a proliferation of alternative wealth and welfare measures in recent decades. The two most famous ones are probably the Human Development Index (HDI), and Gross National Happiness (GNH). We unfortunately have no place to discuss them here. It is clear though that GDP both under- and over-estimates our well-being in crucial ways. With economic and technological growth, tangible products traded at determinate prices on the market become less important. In many ways, our collective infatuation with the GDP is perverse: The tool-set required for GDP calculation was assembled only in the final decades of an Industrial Age obsessed with material necessities, grain yields, concrete slabs and steel tonnage. The question, for our age, then becomes: If ‘the economy’ is not the (weighted) sum-total of material products, what is it?
Figure 3: A Philips Machine, an analogue hydraulic computer used to model income flows within an economy. [Photo courtesy of the Library of the London School of Economics and Political Science.]
Recommended starters on the GDP
Coyle, D. (2015). GDP: A brief but affectionate history. Princeton University Press.
Lepenies, P. (2016). The Power of a Single Number: A Political History of GDP. Columbia University Press.
 Such rankings often involve an adjustment for PPP, that is, the calculations take into account the purchasing power of local currencies.
 Incidentally, besides economics, Goodhart’s Law has been most commonly applied in Education Studies, questioning the validity of the ever-increasing number of performance measures for students, professors, and institutions.
 The economic historian Angus Maddison is credited for launching the gargantuan effort of estimating real GDP per capita numbers all the way to the start of the Common Era. The researchers at the Maddison Project continue his legacy. See Figure 1.
 It is interesting to note that Nazi Germany had no comparable system of national accounts, which has led to disarray in planning, and gave the Western Allies a significant edge. Analyzing this discrepancy, John Kenneth Galbraith argued that ‘Simon Kuznets and his talented people had been the equivalent of several infantry divisions in their contribution to the American war effort’.
 For example, despite all the material destruction, per-capita GDP in Germany was only 10 percent lower in 1945 than in 1938.
 Think of a bank who charges 5$ interest from debtors, and provides 3$ interest to its creditors per year per 100$. It seems that the bank produces -2$ in the process!
 See Figure 3.
About The Author
Zsombor Z. Méder is an economist holding an envy of mathematics and a love of philosophy.