History of Econometrics: From Its Origins to the Present Day

Introduction: what econometrics is and why its history matters

Econometrics is the discipline that brings together economic theory, mathematics, and statistics to measure economic relationships and subject them to empirical testing. The term was coined in 1926 by the Norwegian Ragnar Frisch, but the ambition to “measure” the economy is far older and spans three centuries in the history of ideas.

This article traces that history in two movements. The first follows the evolution of methods, from the political arithmetic of the seventeenth century to today’s machine-learning models; the second describes how the field took root in individual countries, with traditions, schools, and institutions that were often very different from one another.

Understanding this history is useful for anyone using econometrics today. Many current debates—such as the one between structural models and “quasi-experimental” methods, or between interpretability and the predictive power of machine learning—are re-runs of controversies the discipline had already lived through in the 1940s, 1970s, and 1990s.

Origins (17th–19th century): political arithmetic and early measurement

Econometrics was born long before it had a name. In 1690 William Petty, with Political Arithmetick, proposed thinking about the economy “in terms of number, weight, and measure” and was the first to estimate England’s national income. Gregory King (1696) constructed the first empirical demand curve, the so-called “King’s law” on the price of grain.

The eighteenth and nineteenth centuries prepared the statistical tools. Adolphe Quetelet (Belgium, 1835) introduced the “average man” and the application of the normal distribution to social phenomena. Francis Galton (1886) discovered regression toward the mean, Karl Pearson formalized the correlation coefficient (1896), and George Udny Yule applied multiple regression to economic and social problems (1897–1899), laying the foundations for causal analysis via statistical methods.

On the economic front, Augustin Cournot (1838) and Léon Walras (1874) made economic theory mathematical, while William Stanley Jevons (1862–1884) and later Clément Juglar (1862) studied business cycles using historical data. In 1899 Engel formulated his law on household budgets, and in Italy Vilfredo Pareto discovered, in 1896, the income distribution that bears his name.

The birth of the discipline (1900–1945)

Between the two wars, econometrics acquired a name, a scientific society, and a research program. The first decades of the twentieth century saw the pioneering attempts of Henry Ludwell Moore (USA, 1914) to estimate demand curves, Henry Schultz’s first elasticity estimates (1928–1938), and Warren Persons’s business-cycle barometers at Harvard (1919), which sought to forecast the cycle using statistical indices.

The founding moment came on 29 December 1930: in Cleveland, Ragnar Frisch, Irving Fisher, and Charles Roos founded the Econometric Society, “for the advancement of economic theory in its relation to statistics and mathematics.” In 1933 the journal Econometrica was launched, edited by Frisch, and in 1932 the businessman Alfred Cowles financed in Colorado Springs the Cowles Commission for Research in Economics, which moved to Chicago in 1939.

In those same years Jan Tinbergen built the first complete macroeconometric model, first for the Netherlands (1936, 24 equations) and then for the United States on behalf of the League of Nations (1939). The model prompted the famous critique by John Maynard Keynes, who doubted the possibility of deducing stable causal relationships from historical time series: the discipline’s first major methodological debate.

The theoretical turning point came with Trygve Haavelmo. In the essay The Probability Approach in Econometrics (1944) he argues that economic data should be treated as realizations of probabilistic processes, and that models must be formulated and estimated using probability theory. This is the foundation of modern econometric inference and, together with the work of Abraham Wald and Herman Wold on time series, it brings the discipline’s “foundational” phase to a close.

The golden age of structural models (1945–1975)

In the postwar period, the Cowles Commission set the method: simultaneous-equation models grounded in economic theory, identified and estimated with probabilistic techniques. Key contributions include the identification problem (Koopmans, 1949–1950), limited-information maximum likelihood estimators (Anderson and Rubin, 1949), and two-stage least squares (Theil, 1953; Basmann, 1957), later extended to three-stage methods by Zellner and Theil (1962).

Lawrence Klein translated this program into operational models. The Klein–Goldberger model (1955) for the United States and the Wharton Model of the 1960s became the prototype of large-scale macroeconometric models, with hundreds of equations, used by governments and central banks for forecasting and policy simulations. The LINK project (1968) connected national models into a world system.

Meanwhile, the basic tools were maturing. The Durbin–Watson test (1950) for autocorrelation of residuals, the work of Cochrane and Orcutt (1949), and the Chow test (1960) for structural stability entered everyday practice. In the United Kingdom, Richard Stone developed national accounting and demand systems; in the Netherlands, Theil formalized the econometrics of economic policy decision-making.

This is also the phase in which econometrics became institutionalized. Dedicated departments and centers were established (Wharton, Brookings, DRI in the United States; LSE and Cambridge in Great Britain; the Dutch Central Planning Bureau), while the spread of computers made it possible to estimate ever-larger systems.

The methodological revolution (1975-2000)

The 1970s threw the large structural models into crisis. Inflation and stagflation contradicted their forecasts, and in 1976 Robert Lucas formulated his critique: estimated parameters depend on the policies in force and are not invariant, so they cannot be used to simulate new policies. In 1980 Christopher Sims, in Macroeconomics and Reality, attacked the “incredible” restrictions of the Cowles models and proposed VAR models (vector autoregressions), in which all variables are treated symmetrically.

At the same time, time-series econometrics was rebuilt on new foundations. Box and Jenkins (1970) systematized ARIMA models; Granger (1969, 1974) defined causality in a predictive sense and, with Newbold, showed the danger of “spurious regressions”; Dickey and Fuller (1979) provided unit-root tests; Engle and Granger (1987) introduced cointegration, refined by Johansen (1988-1991). Engle (1982) created ARCH models for volatility, extended by Bollerslev (1986) to GARCH, which became the standard in finance.

Microeconometrics made an equally major leap. James Heckman (1976-1979) solved the problem of sample selection; Daniel McFadden (1973-1974) developed discrete-choice models (conditional logit); panel data found their methods with Mundlak, Hausman (1978), Chamberlain, and Arellano–Bond (1991). Lars Peter Hansen (1982) introduced the generalized method of moments (GMM), which makes it possible to estimate dynamic models without fully specifying their distribution. Halbert White (1980) brought heteroskedasticity-robust standard errors into common use.

By the end of the century, the work of Angrist, Card, Krueger, and Imbens on instrumental variables interpreted as “natural experiments” (1990-1996), and the formalization of LATE (Imbens–Angrist, 1994), paved the way for the so-called credibility revolution.

Contemporary econometrics (2000-2026)

The first quarter of the 21st century is dominated by three forces: the “credibility revolution,” big data, and machine learning. The first, dubbed by Angrist and Pischke in 2010, shifts the center of gravity of applied research toward quasi-experimental designs (difference-in-differences, regression discontinuity, instrumental variables, Abadie’s synthetic control) and toward randomized controlled trials, made famous by the work of Banerjee, Duflo, and Kremer on poverty (Nobel 2019). The 2021 Nobels awarded to Card, Angrist, and Imbens seal its success.

In macroeconometrics, DSGE models estimated with Bayesian methods (Smets and Wouters, 2003-2007) become the benchmark tool for central banks, while Bayesian VARs, factor models (Stock and Watson), and nowcasting techniques leverage hundreds of series for real-time forecasting. The 2008 crisis and the 2020 pandemic push the field toward models with nonlinearities, regimes, and tail risk.

From around 2015, the frontier shifts to integration between causal inference and machine learning: double/debiased machine learning (Chernozhukov and coauthors, 2018), causal forests (Athey and Wager, 2018-2019), methods for treatment-effect heterogeneity, and lasso and regularization for high-dimensional control selection (Belloni, Chernozhukov, Hansen). The literature on difference-in-differences with staggered adoption (Callaway-Sant’Anna, Sun-Abraham, de Chaisemartin-D’Haultfœuille, 2020-2021) has corrected decades of applied practice.

The most recent years (2023-2026) show three trends: the use of large language models to construct variables from texts and documents, the growing attention to replicability and pre-registration, and a renewed interest in “micro-founded” structural models estimated with simulation methods, in a synthesis of the two traditions that have divided the discipline for half a century.

Interactive timeline: Nobel Prizes and foundational contributions

The awards come in three waves: the European founders (1969-1989), the generation of the methodological revolution (2000-2013), and the generation of causal inference (2015-2021). Hovering over each point shows the prize citation; the color indicates country of birth.

Many decisive contributions have not received the Nobel. The table below lists a few that every textbook considers foundational.

United States

The United States is the country where econometrics became institutionalized and where, since the postwar period, most research has been produced. The pioneers were Henry L. Moore at Columbia (1908-1929), his student Henry Schultz at Chicago, and Irving Fisher at Yale, co-founder of the Econometric Society. Wesley Mitchell’s NBER (1920) and the Harvard barometer represent the “empiricist” tradition that Koopmans would attack in 1947 as “measurement without theory.”

The Cowles Commission (Chicago, 1939-1955; later at Yale as the Cowles Foundation) was the true heart of the field in the 1940s: Koopmans, Marschak, Haavelmo, Klein, Arrow, Hurwicz, and Simon developed the simultaneous-equations method there. In the 1960s, leadership shifted to large-scale models (Klein at Wharton, Brookings, Modigliani at MIT-FRB) and to the econometrics of firms and households (Griliches at Harvard, Jorgenson).

Since the 1970s, American universities have led every turning point: Chicago (Lucas, Hansen, Heckman), Minnesota (Sims, Sargent), Berkeley (McFadden), MIT (Hausman, Angrist, Duflo), Stanford (Imbens, Athey), Princeton (Card, Krueger). The United States holds the overwhelming majority of Nobel prizes awarded for econometric contributions and hosts the most influential journals and societies.

United Kingdom

The United Kingdom is the cradle of modern statistics and of an econometric tradition autonomous from the American one. Galton, Pearson, Yule, and Fisher (the latter at Rothamsted, with experimental design and maximum likelihood) provided the basic tools between 1886 and 1930. In the 1930s, by contrast, Keynes and Cambridge were skeptical of econometrics, as the dispute with Tinbergen shows.

In the postwar period, two schools emerged. At Cambridge, Richard Stone (Nobel 1984) built national accounts and demand systems, and the Department of Applied Economics trained generations of econometricians. At the London School of Economics, Denis Sargan, Jim Durbin, and later David Hendry developed in the 1960s–1970s the “LSE methodology”: specification tests, general-to-specific modeling, and—with Hendry—a dynamic modeling approach that culminated in the cointegration theory developed at Oxford and San Diego with Granger and Engle.

Clive Granger, trained at Nottingham, received the Nobel in 2003 for cointegration. Today the main centers are Oxford, Cambridge, LSE, UCL (with Chernozhukov in part and Blundell’s structural microeconometrics group), and Warwick; the Institute for Fiscal Studies is Europe’s most important center for econometrics applied to public policy.

The Netherlands and Scandinavia

The Netherlands, Norway, Sweden, and Denmark carry a weight in the history of econometrics disproportionate to their size: three of the discipline’s founders (Frisch, Tinbergen, Haavelmo) are Norwegian or Dutch, and the first two received the first Nobel Prize in Economics in 1969.

Netherlands. In 1936 Jan Tinbergen builds the first national macroeconometric model and in 1945 founds the Central Planning Bureau (CPB), the first institution to use econometric models for economic policy. Henri Theil, at the Netherlands Econometric Institute in Rotterdam (1956), develops two-stage least squares, the theory of economic-policy decision-making, and the Rotterdam demand system. Rotterdam’s degree program in “econometrics” (1956) is the first in the world to bear that name; today Tilburg, Amsterdam, and Groningen remain top-tier centers.

Norway. Ragnar Frisch coins the term, founds the Society, edits Econometrica for more than twenty years, and at the University of Oslo creates a school that includes Trygve Haavelmo, whose 1944 essay lays the foundations of the probabilistic approach. Both receive the Nobel Prize (1969 and 1989). Statistics Norway and Norges Bank continue this tradition with the MODAG and NEMO models.

Sweden and Denmark. In Sweden, Herman Wold (Uppsala) develops in the 1930s–1950s the theory of stationary time series (Wold decomposition) and recursive models, in critical dialogue with Cowles. In Denmark, Søren Johansen (Copenhagen) formulates in 1988–1991 the maximum-likelihood procedure for cointegration, which has become a global standard, and Katarina Juselius develops its applications.

France

France has a very long-standing mathematical tradition in economics (Cournot in 1838, Walras, Dupuit) but an econometrics that developed mainly within the state. In the postwar period, INSEE (1946), ENSAE, and the Grands Corps formed a school of “engineer-economists”: Edmond Malinvaud, director of INSEE, published in 1964 the Méthodes statistiques de l'économétrie, for decades the leading European reference textbook. Maurice Allais (Nobel 1988) and Marcel Boiteux represent the more theoretical strand.

In the 1970s–1980s, CEPREMAP and CREST developed macroeconometric models for planning (DMS, METRIC) and structural microeconometrics. Over the past thirty years, the French school has produced major contributions in asymptotic theory and simulation methods (Christian Gouriéroux and Alain Monfort, with indirect inference methods, 1993), in panel econometrics and policy evaluation (Bruno Crépon, Xavier D'Haultfœuille, co-author of modern DiD estimators), while the Toulouse School of Economics and the Paris School of Economics are among Europe’s leading centers. Thomas Piketty has made the historical econometrics of incomes and wealth widely known.

Germany

Nineteenth-century Germany was the birthplace of social statistics (Engel, Lexis, Bortkiewicz) and of the historical school, which, however, was wary of mathematical theory: a legacy that, together with the diaspora of the 1930s (Marschak, Wald, Hurwicz, Morgenstern left the German-speaking countries), delayed the emergence of an econometric school. As early as 1935, Ernst Wagemann at Berlin’s Institut für Konjunkturforschung was already using statistical methods for business-cycle analysis.

The postwar revival ran through economic research institutes—DIW (Berlin), Ifo (Munich), IfW (Kiel), RWI and ZEW—and through the Bundesbank, which since 1970 has used a quarterly macroeconometric model. In academia, Wilhelm Krelle (Bonn) built in the 1960s the first major German macroeconometric model, and Heinz König (Mannheim) established the tradition of applied microeconometrics.

Today Germany is Europe’s second pole after the United Kingdom: Mannheim, Bonn, LMU Munich, Humboldt University of Berlin and Konstanz in theory and time series (Helmut Lütkepohl, author of the standard reference manual on VARs, Jörg Breitung, Joachim Grammig), and the IZA (Bonn) and DIW institutes in labor microeconometrics and policy evaluation. The journal Empirical Economics and the Deutsche Bundesbank Research Centre are the main dissemination channels.

Italy

Italy boasts a first-rate statistical-economic tradition between the late nineteenth and early twentieth centuries: in 1896 Pareto discovered the income distribution; in 1912 Corrado Gini introduced the concentration index; in 1907 Rodolfo Benini applied multiple regression to demographic and economic problems; in the 1930s Bruno de Finetti founded subjective probability, now underpinning Bayesian econometrics. The Società Italiana di Statistica (1939) and Istat (1926) are among the foundational institutions.

Econometrics in the strict sense arrived later, and mainly through the Banca d’Italia, which in the 1960s built the M1BI model (1970) with Franco Modigliani, Antonio Fazio and Franco Cotula, and then the Quarterly Model (1986), still updated today. In universities, Giorgio Fuà (Ancona), Luigi Pasinetti and the Prometeia group (Bologna, 1974) developed applied models; Carlo Giannini and Marco Lippi contributed to the theory of structural VARs and factor models.

From the 1990s onward, Italian econometrics became increasingly international: Bocconi, the European University Institute in Fiesole (where Hendry and Favero taught), Bologna, Padua, Rome Tor Vergata, and EIEF have trained researchers who work at the world’s leading universities (including Fabio Canova, Carlo Favero, Luca Benati, Francesca Molinari, Andrea Ichino, Enrico Moretti). The Italian Econometric Society (SIdE), founded in 2003, now brings the community together and organizes doctoral schools of international standing.

Canada, Australia and other Western countries

Canada. Canadian econometrics grew in the postwar period around the Bank of Canada (the RDX1 model, 1969, and RDX2, 1972, among the most advanced of their time) and the universities of Toronto, Western Ontario, and Queen's. In theory, James MacKinnon and Russell Davidson stand out (bootstrap methods, specification tests, 1993 textbook), as do Angelo Melino and John Galbraith and, in microeconometrics, David Card (Canadian, Nobel 2021) and Thomas Lemieux. Montréal (CIREQ) is an important hub for Bayesian methods and time series (Jean-Marie Dufour, Éric Renault).

Australia. It emerged around Monash, ANU, and Sydney in the 1960s–1970s with computable general equilibrium models (ORANI, 1977, by Peter Dixon) and the Reserve Bank’s macroeconometric models. Peter Phillips, an Australian who moved to Yale, is one of the leading theorists of non-stationary time series; Adrian Pagan contributed to test theory and macro modeling; Monash remains a reference center for forecasting (Rob Hyndman).

Switzerland, Belgium, Spain. Geneva hosted the League of Nations and Tinbergen’s work; Louvain (CORE, 1966) trained generations of European econometricians, with Jacques Drèze, Jean-François Richard, and Luc Bauwens in Bayesian methods; Spain, with CEMFI (Madrid, 1987) and Manuel Arellano, made decisive contributions to panel econometrics.

Japan

Japan imported econometrics in the postwar period, but quickly became an original producer. In the 1950s–1960s the Economic Planning Agency and the Bank of Japan built the first planning models, while the first generation took shape at the University of Tokyo and Hitotsubashi (Shinichi Ichimura, Michio Morishima, Takeshi Amemiya). Many Japanese econometricians later worked in the United States: Amemiya (Stanford) wrote a landmark textbook in 1985 and contributed to models with limited dependent variables.

The Japanese Economic Association, the Institute of Statistical Mathematics, and the Japan Statistical Society supported a strong school in asymptotic theory and time series. In 1973 Hirotugu Akaike formulated the AIC information criterion, one of the most widely used tools in the world for model selection. Fumio Hayashi (Tokyo) authored the textbook Econometrics (2000), adopted in many PhD programs; Nobuhiko Kiyotaki, Kazuo Ueda, and Taisuke Nakata link macroeconomic theory and empirical evidence.

Today the main centers are the universities of Tokyo, Kyoto, Hitotsubashi, and Osaka and the Bank of Japan’s Institute for Monetary and Economic Studies (IMES), with particular strength in macroeconometric modeling of zero rates and deflation—a natural laboratory that anticipated the problems faced by other advanced economies after 2008.

China: from imported econometrics to autonomous production

In China, econometrics was absent until the 1978 reforms: in the planned economy, statistics served plan accounting, not hypothesis testing. The turning point came in 1980, when Lawrence Klein led a seven-week summer seminar in Beijing, organized by the Chinese Academy of Social Sciences, which trained the first generation of Chinese econometricians (including Gregory Chow, who in subsequent years served as a bridge between Princeton and Chinese universities). In 1982 the first econometrics textbook in Chinese was published, and in 1984 the Chinese Association of Quantitative Economics was founded.

Between 1990 and 2010 the field consolidated through overseas training and the return of scholars: Justin Yifu Lin, Cheng Hsiao (textbook on panel data, 1986), Jushan Bai (time series and factor models), Yongmiao Hong (specification tests and financial series). Economics schools emerged in Beijing (CCER, 1994, later NSD), Tsinghua, Fudan, Xiamen (WISE, 2005), and the journal China Economic Review. The Chinese Economists Society and the Econometric Society have held regular conferences in China since 2006.

Today China is the world’s second-largest producer of econometrics research by number of articles, with strengths in high-dimensional time series, financial econometrics, large panel datasets, and policy evaluation using administrative data; the quality of official statistics, however, remains a widely debated constraint. The handling of Chinese “big data” (platforms, payments, registries) represents one of the most innovative frontiers of the 2020s.

The major emerging countries: India, Brazil, Russia, and others

India. It is the emerging country with the oldest tradition. In 1931 Prasanta Chandra Mahalanobis founded the Indian Statistical Institute in Calcutta, introduced the distance that bears his name, and designed the Second Five-Year Plan (1956) with a two-sector model; C.R. Rao, trained there, is one of the greatest statisticians of the twentieth century (Cramér–Rao bound, 1945). The Delhi School of Economics, Mumbai’s IGIDR, and ISI now host a strong school in development econometrics; the Indian diaspora (Amartya Sen, Abhijit Banerjee, Raj Chetty, Rohini Pande, Arvind Subramanian) is central to global research.

Brazil. Brazilian econometrics emerged in the 1960s–1970s with Fundação Getúlio Vargas, IPEA, and the University of São Paulo, and was strengthened by the 1994 stabilization and by Central Bank research. Marcelo Fernandes, Marcelo Medeiros, and PUC-Rio are reference points in financial econometrics and time series; the breadth of Brazilian administrative records (RAIS) has made the country a laboratory for labor microeconometrics and the evaluation of social programs (Bolsa Família).

Russia. The USSR had a great mathematical school (Kantorovich, Nobel 1975; Slutsky; Kolmogorov) but an econometrics hampered by ideology; in 1927 Eugen Slutsky discovered that the sum of random variables generates apparent cycles, a fundamental result for time series. After 1991, the New Economic School (Moscow, 1992) and the Higher School of Economics created a modern community, now partly dispersed abroad.

Other countries. Mexico (ITAM, Colegio de México), Turkey (Bilkent, Koç), South Korea (KDI, Seoul National), South Africa (Stellenbosch), and Indonesia have developed national schools, often around central banks and research centers focused on development programs; Korea in particular is now among the top ten countries in the world for econometrics output.

Comparative tables and charts: schools, methods, and countries’ weight

The chart counts institutions by country in the RePEc econometrics ranking, which measures the output of registered researchers in the field. The United States dominates, the United Kingdom is the second hub, and among mid-sized countries Australia, Canada, the Netherlands, Spain, and Denmark stand out; China appears with three institutions, Italy with two (Bocconi and EIEF). The ranking understates countries whose researchers work abroad, such as Italy and India.

The major schools compared

National vocations

Conclusions and outlook

The history of econometrics is the story of a pendulum swinging between theory and data. The Cowles Commission put theory at the center; the Lucas critique and Sims’s VARs shifted the weight toward data; the credibility revolution shifted it again, toward research design; causal machine learning is now seeking a synthesis, in which statistical flexibility is used to identify effects that are economically interpretable.

Geographically, the discipline was born in Europe (Norway, the Netherlands, the United Kingdom), grew in the United States—which has remained its center since 1945—and spread to Asia starting in the 1980s. China and India are now leading producers, and over the next decade the availability of administrative and platform data in emerging countries could push the empirical frontier further outside the West.

Three open questions will define the coming years: how to reconcile interpretability and predictive power in machine-learning models; how to ensure replicability and transparency in an era of proprietary data; and how to bring economic theory back into methods that, in reaction to structuralist excesses, long kept it at the margins.

Why it matters

Econometrics is the tool central banks use to estimate expected inflation, governments use to assess the effects of a reform, and courts use to quantify damages. Knowing its history helps one read with the right skepticism the numbers that politics cites every day: behind every estimate there is a school, an assumption, and often a controversy. And the geography of the discipline—with half of the leading institutions in the United States and China rising rapidly—says a lot about who will define the methods by which the economy of the coming decades will be measured.

Essential bibliography

• Mary S. Morgan, The History of Econometric Ideas, Cambridge University Press, 1990.

• Roy J. Epstein, A History of Econometrics, North-Holland, 1987.

• Duo Qin, The Formation of Econometrics: A Historical Perspective, Oxford University Press, 1993.

• Duo Qin, A History of Econometrics: The Reformation from the 1970s, Oxford University Press, 2013.

• David F. Hendry and Mary S. Morgan (eds.), The Foundations of Econometric Analysis, Cambridge University Press, 1995.

• Trygve Haavelmo, “The Probability Approach in Econometrics,” Econometrica, 12 (supplement), 1944.

• Christopher A. Sims, “Macroeconomics and Reality,” Econometrica, 48, 1980.

• Joshua D. Angrist and Jörn-Steffen Pischke, “The Credibility Revolution in Empirical Economics,” Journal of Economic Perspectives, 24(2), 2010.

• Susan Athey and Guido W. Imbens, “Machine Learning Methods That Economists Should Know About,” Annual Review of Economics, 11, 2019.

• Gregory C. Chow, “Teaching Economics and Studying Economic Reform in China,” China Economic Review, 2000.

• Carlo Giannini and Marco Lippi, contributions on structural VARs and factor models; SIdE, Storia e attività della Società Italiana di Econometria.

The dates and attributions reported in the article derive from the historical literature cited above and should be regarded as approximate where the primary source has not been consulted directly.