Raj Chetty and the New Scientism: Big Data, Economic Engineering, and the Failure of Economic Education
By Nikolai Wenzel
Reveries of a Solitary Walker
I confess that I am an atypical tourist. When last in Paris, I eschewed the charred Notre Dame cathedral and the urban oasis of the Luxembourg Gardens. Instead, I headed to the Citéco1 Museum of the Economy. The museum is quite well done, with foundational exhibits on exchange, individual autarky, and money, before macroeconomic debates and policy questions. The imagined debate between Milton Friedman and John Maynard Keynes is the most insightful, even if it pales compared to the Keynes-Hayek rap.2 The chart of economists from Adam Smith to the present is largely comprehensive, as it includes both Karl Marx and the Austrians, and both socialists and new institutionalists. The glaring omission is Public Choice theory, a school that could have informed the exhibits on externalities and public policy. It was fun to play social engineer on the French retirement machine: I had the power to tinker with the retirement age and the (national and mandatory) retirement contributions. If I pushed the retirement age too high, workers would frown (and eventually go on strike, because they are French). If I pushed the contributions too low, the retirees would frown (and, because it’s France, workers would also go on strike—not for their own low contributions, but to raise the contributions of others!). My job, according to the instructions, was to find the socially optimal combination of retirement age and retirement tax.
A few years ago, I strayed from London’s beaten path. Instead of visiting Buckingham Palace or Abbey Road studios, I headed to the Museum of Science and Industry, in quest of the Phillips Machine.3 The machine was invented by economist Alban William “Bill” Phillips (of the Phillips curve). The machine demonstrates Keynesian hydraulics, with dials, and tubes filled with colored water. The economic engineer can play with taxes, government spending, and the money supply, to affect GDP (represented by the level of colored water in a tank).
There’s also an exhibit at the Chicago Fed that allows tourists to set the target federal funds rate and the base year, then watch inflation….
I probably should not confess the glee I got from leaving behind the Louvre, the Tate, and the Field, in favor of economic tinkering. But the real rush came from the sensation of power—I can’t make my students read their assignments, but, for a moment, I could guide the economy towards a socially optimal equilibrium!
Alas, my idols were clay. “Socially optimal” retirement policies have led to strikes in France, with almost a million people demonstrating in the streets of Paris in early December—and the bankruptcy of U.S. Social Security. Keynesian interventionism causes stagflation, perpetual debt, and government growth. And the Federal Reserve’s monetary policy has caused boom and bust cycles, like the Great Financial Crisis of 2008.
Despite that knowledge, I have jumped with delight into Raj’s Chetty’s Harvard course, ECON 1152: Using Big Data to Solve Economic and Social Programs.4 I am already thinking of ways to incorporate some of his material into my courses (as he explicitly encourages). And I will use some of Chetty’s findings in my work on poverty.
Chetty’s overall research project uses Big Data to solve economic and social problems. In his course, he adopts a direct approach, over the theory and “abstract methods” of introductory economics. He proposes to generate excitement first, then encourage students to study theory.
The course introduces the students to Big Data analysis and application to social problems over 18 lectures: (1) equality of opportunity, social mobility, income inequality, and education; (2) racial disparities, health, criminal justice, climate, tax policy, and international development; with (3) a bit of theory on data-driven behavioral economics and framing.
Chetty is an exciting lecturer and an innovative thinker. He is addressing a real problem in the literature (understanding social problems) and a real problem in education (flawed principles of economics classes). But while I understand the appeal of Chetty’s Big Data approach, it has deep problems of its own. I cannot forget the lessons of all that failed economic tinkering.
The Debate Is Dead—Long Live the Debate!
In many ways, Chetty is just another applied mathematician playing at economic engineering. But his philosophy of moving towards “natural science data” over “speculation” is somewhat radical. Advances in Big Data and AI mean not only that more data is available (including proxies for emotions, such as clicks or Facebook/Twitter comments), but also that social scientists can now “approximate the natural sciences.”
The skeptic must ask two questions: (1) how Big Data fits in within the greater debate of scientism; and (2) whether Big Data will now win the socialist calculation debate. These questions lead, naturally, to F.A. Hayek. After cutting his teeth against the positivists of early 20th-century Vienna, Hayek turned to the problem of knowledge. While the main implications of his thinking were institutional, his methodology is largely captured in The Counter-Revolution of Science: Studies on the Abuse of Reason (“CRS”).5
Hayek worried about the “tyranny of the ‘scientific’ method” (CRS 20), dismissing its effectiveness in economics and social theory: “although in the 120 years or so during which this ambition to imitate science in its methods rather than its spirit has now dominated social studies, it has contributed scarcely anything to our understanding of social phenomena” (CRS 21). This worry is echoed in Buchanan’s lament:6
- If not an economist, what am I? An outdated freak whose functional role in the general scheme of things has passed into history? Perhaps I should accept such an assessment, retire gracefully, and, with alcoholic breath, hoe my cabbages. Perhaps I could do so if the modern technicians had indeed produced “better” economic mousetraps. Instead of evidence of progress, however, I see a continuing erosion of the intellectual (and social) capital that was accumulated by “political economy” in its finest hours.
The problem, explained Hayek, is that the natural sciences and the social sciences are inherently different. The facts of the social sciences are the beliefs held by actors. These opinions can be true or false—what is relevant is that they motivate behavior. Social and economic outcomes are the “the result of human action but not human design.” Society can be understood only via individual actions, and spontaneous collaboration yields things greater than any individual mind can comprehend. This explains the importance of complex emergent phenomena that aggregate information and coordinate action—from the market to norms, and from emergent law to institutional correction. Hayek explains that “information-gathering mechanisms such as the market enable us to use such dispersed and unsurveyable knowledge to form super-individual patterns” (The Fatal Conceit,7 “TFC,” 15.)
The natural sciences are well equipped to observe, analyze, and explain simple phenomena. But individuals are not atoms, and their choices affect the choices of others: “the individuals are merely the foci in the network of relations and it is the various attitudes of the individuals towards each other… which form the recurrent, recognizable and familiar elements of the [social, economic, or institutional] structure” (CRS 59).
Hayek continues by explaining that statistics can tell us about individuals as building blocks of social structures, but “no statistical element about the elements can explain to us the properties of the connected wholes [such as prices or language]” (CRS108-109). Statistics can give us good information at a particular time and place, and provide practical applications for theories, but only for a snapshot, and not overall patterns (CRS110). Even though statistics can show patterns, “unless we can understand what the acting people mean by their actions, any attempt to explain them, that is, to subsume them under rules which connect similar situations with similar actions, is bound to fail” (CRS 53).
This application of the natural sciences to social phenomena is part and parcel of a philosophical current, which Hayek labels “false individualism.” This strain (associated with Rousseau, positivism, socialism, and communism) is the “product of an exaggerated belief in the powers of individual reason and of a consequent contempt for anything which has not been consciously designed by it or is not fully intelligible to it” (“Individualism: True and False,”8 “ITF,” 8). Hayek later dubbed this notion, that the human mind can engineer society with better outcomes than spontaneous order, “the fatal conceit.” Naturally, scientism has had pernicious effects. Think of bankrupt pension funds, permanent deficit, and boom and bust cycles.
But what about Big Data and AI? The famous “calculation debate” embodied the conflicting views of society as the result of human action versus human design. As early as 1920, Ludwig von Mises explained that full central planning could not function. Without private property, there are no prices, and thus no competition. Without the profit and loss system to convert individual choices into efficient outcomes, according to division of labor and knowledge, there cannot be rational allocation of scarce resources. Critics of the market argued that central planning could in fact work, as central planners found prices through a process of trial and error.
Hayek cautioned would-be central planners that “the number of separate variables which in any particular social phenomenon will determine the result of a given change will as a rule be far too large for any human mind to master and manipulate them effectively” (CRS 73). Might Big Data and AI overcome the limitations of the human mind, so that central planning is now possible? No—but why? First, as explained above, acting humans are not predictable atoms. Second, there is more to spontaneous order than merely aggregation of information. Prices provide information and coordination, but they also provide incentives; the market mechanism rests on property prices to promote socially useful behavior (ITF 12).
Chetty’s approach holds much promise—if it follows two conditions: first, that Big Data be combined with sound economic theory; and second, that social scientists act very carefully before they use Big Data to act as economic engineers. They would indeed do well to remember Hayek’s warning that it “the curious task of economics is to demonstrate to men how little they know about the systems they purport to design” (TFC, 76). This is also exemplified in a rich literature on expert failure and the tyranny of experts.
Torturing the Data and Engineering Society
Chetty emphasizes social mobility instead of poverty. He maps mobility by neighborhood, at the national level. He then suggests policy solutions that range from vouchers for families to move from low- to high-mobility areas, and (in a rare nod to government failure), changes in zoning regulations that lower housing opportunities, along with local initiatives, including better schools and mixed-income housing. This is promising. But there is still something important to be said for examining the causes of poverty. Indeed, the U.S. War on Poverty has been an abysmal failure, as poverty rates have stayed the same despite massive government spending. It would be useful to discuss the 25% of Americans who require a license to practice a job… the regressive effects of regulation… over- and under-policing… or the 10% of GDP that is diverted to compliance with federal regulations.
Chetty analyzes the share of income “going to” the top 1%, as if it were either stolen, or magically allocated. Barring a discussion of cronyism (which is absent), it would be more appropriate to refer to income “earned by” the top 1%—then ask questions about productivity at the bottom (rather than redistribution from the top). Chetty asks his students, “How would you be doing if inequality had not increased,” then contends the following: “A household making $25,800 today would instead be making $35,200 if inequality had not changed since 1980. In other words, if growth had been evenly shared, this household would have earned 37% more.” But do we know this? Is growth thus linked to distribution? Without innovation (and early gains at the top), would there have been the same level of growth? Would the $25,800 household even be earning that much? Maybe. But he assumes that the pie would have increased, regardless of details.
Chetty examines both government and (mixed-) market approaches to education. The record on charter schools is mixed (perhaps a small wonder, if one considers that they are still under the thumb of educrats). What is more, Chetty worries that parents may not make well-informed choices; but who would make better choices? Surely not the educrats who have mucked up education. Are parents not making poor choices because their options are so limited? On government schools, Chetty’s data suggests that smaller class sizes and more effective teachers have a huge positive impact on educational outcomes. The policy implication, of course, is to fund smaller classes and promote teachers based on merit. However, given the current government monopoly on funding education and quasi-monopoly on its provision, it’s hard to see how these could work outside some fundamental changes—which require deep imagination rather than rearranging the deck chairs on the Titanic. It is no surprise that government spending per student has increased over the past 50 years, while test scores have stayed flat (and maybe even decreased if inflation in standardized tests matches inflation in K-12 and college grades).
On health insurance, Chetty concludes from Big Data that better health insurance leads to better health outcomes. He concludes that government intervention in health insurance is critical, for two reasons: first, lower-income individuals won’t buy it unless it is subsidized or provided by the government; and second, adverse selection would lead to a market collapse. Health insurance in the US is broken, no doubt. But there is no discussion about a century of interventions. The AMA has restricted the supply of doctors; regulations on health insurance sales have increased prices; tax exemptions for employers have decreased portability and competition; and subsidies (Medicare, Medicaid, etc.) have increased price. Consider that before the Patient Protection and Affordable Care Act, fully half of all health insurance was already provided by state or federal sources. Instead of advocating further intervention, Chetty would do well to consider the benefits of a free market supplemented by civil society.
And the list goes on.
Abstract Theory, Statistics, and Plausible Stories
Chetty is right that the teaching of principles of economics is broken. But he adopts the wrong remedy. Instead of “abstract methods” and theory, he recommends the answers and evidence from data. Both are misguided.
“What Should Economists Do? An Appreciation”, by Donald J. Boudreaux. Book review of James Buchanan’s book. Library of Economics and Liberty, Mar. 4, 2019. See also the EconTalk podcast episodes Susan Athey on Machine Learning, Big Data, and Causation, and Binyamin Appelbaum on the Economists’ Hour.
Although Chetty explicitly states that Big Data is “intended to complement traditional Principles of Economics (Econ 101) courses,” how many students will really take econ 101 after Chetty’s engaging approach? More important, Chetty’s course arguably contains very little economics; it does make brief allusion to incentives, supply and demand, and market equilibrium. But it is basically an applied statistics class.
Statistics are no substitute for economic thinking, and I fear Chetty may be doing his students a disservice by introducing them to Big Data too early. For all its complexity, the economic way of thinking really comes down to two fundamental insights, both of which are entirely absent from Chetty’s course. First, he does not discuss scarcity (and thus opportunity cost, cost-benefit analysis, and division of labor and knowledge according to comparative advantage). Second, he leaves out the price mechanism, with its ability to provide incentives, aggregate knowledge, coordinate preferences, guide the self-love of the baker, butcher and brewer into our own dinners, and generally maximize the allocation of scarce resources among competing wants—but also the unintended consequences of market interventions, from price controls and subsidies to regulations.
The debate over market failure is still open; but premature introduction to statistics, without a solid foundation in scarcity and market (dis-)equilibrium is likely to turn bright and eager students into hubristic economic engineers who will play with the economy to achieve their visions of social justice. In the words of Adam Smith:
- The man of system… is apt to be very wise in his own conceit, and is often so enamoured with the supposed beauty of his own ideal plan of government, that he cannot suffer the smallest deviation from any part of it. He goes on to establish it completely and in all its parts, without any regard either to the great interests, or to the strong opposition which may oppose it: he seems to imagine that he can arrange the different members of a great society, with as much ease as the hand arranges the different pieces upon a chess board: he does not consider that the pieces upon the chess board have no other principle of motion besides that which the hand impresses upon them; but that, in the great chess board of human society, every single piece has as principle of motion of its own, altogether different from that which the legislature might choose to impress upon it. If those two principles coincide and act in the same direction, the game of human society will go on easily and harmoniously, and is very likely to be happy and successful. If they are opposite or different, the game will go on miserably, and the society must be, at all times, in the highest degree of disorder. (Theory of Moral Sentiments9)
Like a sorcerer’s apprentice, Chetty wields Big Data and AI with great promise—and great peril.
 Find the course online at OpportunityInsights.org: Using Big Data to Solve Economic and Social Problems.
 Studies on the Abuse and Decline of Reason: Text and Documents, by F. A. Hayek. Available from Liberty Fund Books.
 What Should Economists Do? by James M. Buchanan. Preface by Geoffrey Brennan and Robert D. Tollison. Liberty Fund, Inc., November 1979. Sixteen of Buchanan’s essays collected in book format. Available through the Liberty Fund Book Catalog.
*Nikolai G. Wenzel is the L.V. Hackley Chair for the Study of Capitalism and Free Enterprise, and Distinguished Professor of Economics, Broadwell College of Business and Economics, Fayetteville State University (Fayetteville, NC).