Few economists have had as broad and lasting an impact on academic research and economic policy as Chris Sims. His work set the standard for how theory and data should interact. In this post, we review his key contributions and their influence on the field. All four of us were Chris’s doctoral students – Frank and Marco at Yale, Francesco and Giorgio at Princeton. As a result, this review is subjective and includes a few personal anecdotes.

Chris’ web page lists his papers by topic. Tellingly, he starts with “Macroeconomics”. This is not an accident. Despite his immense contributions to econometrics, Chris thought of himself first and foremost as a macroeconomist. His students can be placed along a spectrum between pure macroeconomics and pure econometrics – some closer to one of the ends, many somewhere in between. Chris’s contributions spanned this entire spectrum, unified by a single goal: to deepen our understanding of the macroeconomy.

Chris influenced modern macroeconomics by stressing the importance of confronting theories with data using formal econometrics methods, and the need for a joint, system-wide analysis of macroeconomic dynamics, as opposed to one-equation, one-impulse-response, one-variable-at-a-time approaches. Policy analysis at central banks and similar institutions demands that these prescriptions are met. While important insights can be gained from qualitative models, policy decisions are ultimately quantitative (e.g. cut by 25 or 50 basis points) and ask for the support of quantitative, system-wide approaches. Policy institutions using models for macroeconomic analysis, such as VARs, factor models, estimated DSGE models, owe much to Chris’s influence. In addition, Chris made seminal contributions to specific strands of literature. In this column, we review six of them – arguably the most important: vector autoregressions (VARs), the promotion of Bayesian methods in macroeconometrics and DSGE models in central banks, factor models, the Fiscal Theory of the Price Level, and rational inattention. We begin with the work for which he was awarded the Nobel Prize.

Macroeconomics and reality (Sims 1980): The Wall Street Journal obituary for Chris had an apt title: “Christopher Sims, Economist Who Taught the Data to Speak.” VARs indeed taught the data to speak to economists about macroeconomic dynamics, and the language was that of impulse response functions. 

Before Chris, times series econometrics largely consisted of Box-Jenkins-type analysis of univariate autoregressive integrated moving average (ARIMA). This literature made progress in forecasting (AR models are still hard to beat) but was less useful to macroeconomists interested in modelling the joint behaviour of multiple variables to answer questions such as what happens to output and inflation when the central bank raises interest rates. This set of questions was addressed by so-called Cowles Foundation models, which were simultaneous equations models in the tradition of Haavelmo, Koopmans, Klein, and many others. These models consisted of accounting identities and behavioural equations (e.g. consumption, investment, …), each identified through exclusion restrictions and/or exogeneity assumptions. Chris famously described these restrictions as “incredible”. The macroeconomics emerging at the time, based on microfoundations and general equilibrium theory, had little use for Cowles Foundation models (see Tom Sargent’s “cross- equation restrictions’’; Sargent 1981), and needed  new empirical tools.

VARs were the answer. Their flexibility made them potentially consistent with a wide range of theories, and able to inform theory by describing the dynamics in the data. To be clear, Chris did not ‘invent’ VARs – they are simply multivariate extensions of AR models. His key innovation was to link VAR forecast errors to economic shocks, such as unanticipated changes in monetary policy, and to use VARs to trace out the dynamic response of the macroeconomy to those shocks.  In this sense, impulse responses ‘let the data speak’ about macroeconomic dynamics.

Minnesota priors and Bayesian econometrics: VARs are extremely flexible statistical tools, but they involve many parameters, which led Chris to Bayesian methods to address this challenge. The Minnesota priors he developed with Tom Doan and Bob Litterman (Doan et al. 2007) have stood the test of time and are widely used to this day, perhaps more than ever. Decades before ‘regularisation’ became a buzzword in econometrics, statistics, and machine learning, Chris emphasised the need for incorporating additional information – or ‘prior distributions’, in the language of Bayesian inference – when estimating models with limited data, such as cases with only a handful of observations per parameter. His advocacy of Bayesian methods extended well beyond VARs. He frequently gave lectures on econometric inference, and the title of one of his presentations succinctly captures his perspective: “Bayesian Methods in Applied Econometrics, or, Why Econometrics Should Always and Everywhere Be Bayesian” (Sims 2007). Chris has been very influential in this respect and many empirical macroeconomists, definitely including the authors of this post, nowadays use Bayesian approaches.

The use of DSGE Models at central banks: If one divides modern macroeconometrics – that is, the development and application of rigorous econometric techniques to macroeconomic questions – into semi-structural approaches (VARs and dynamic factor models) and fully structural approaches (dynamic stochastic general equilibrium, or DSGE, models), then Chris’s name is most closely associated with the (structural) VAR branch. However, he was also instrumental in promoting the use of DSGE models at central banks. His 1994 paper with Eric Leeper, “Toward a Modern Macroeconomic Model Usable for Policy Analysis,” sets the agenda for moving toward models that are both estimated using likelihood-based methods and incorporate a meaningful role for monetary policy. This agenda gained traction about a decade later when DSGE models took off. On a personal note, when two of the authors of this column were writing their dissertations at Yale in the mid-1990s, Chris championed the term DSGE, as he thought that there was more to dynamic macro than just the real business cycle models popular at that time. As his students, we took this terminology for granted, but were often met with bewilderment outside of Chris’ office. It took a few more years for the profession to adopt it as standard.

Factor models (Sargent and Sims 1977): In an office interaction, one of the authors of this post early in his PhD tried to explain to Chris what factor models are. Chris, with a very unassuming tone, interrupted him: “I have done some work on factor models myself, even if not many people are aware of that!” In fact, Tom Sargent and Chris, contemporaneously with John Geweke, who was their student at Minnesota, introduced factor models to macroeconometrics in the late 1970s – they called them “index models.” This contribution is little known as Chris soon dropped factor models in favour of VARs, and it took more than a decade for these tools to become established in the profession.

The fiscal theory of the price level (Sims 1994): Chris also played a central role in the development of the fiscal theory of the price level (FTPL). He recalls becoming interested in the topic after attending a seminar given by Mike Woodford at Minnesota. He found the idea that inflation is ultimately a fiscal phenomenon, determined by the interaction between monetary and fiscal policy, deeply compelling. While the FTPL initially encountered significant resistance within the profession, it has gained substantial traction in recent years, once again demonstrating Chris’s ability to lead the way. In 2016, Chris addressed central bankers at the Jackson Hole Economic Symposium (Sims 2016), where he argued that fiscal deficits could substitute for ineffective monetary policy, provided that the deficits are “[…] seen as financed by future inflation, not future taxes or spending cuts.” According to several economists, the post-COVID inflation episode can be interpreted as a validation of Chris’s ideas.

Implications of rational inattention (Sims 2003): Chris’ work on rational inattention (RI) seems to come out of the blue in terms of his body of work up to that point. In fact, Chris’ undergraduate dissertation was on information theory and he evidently kept thinking about this topic. In the 1998 Carnegie-Rochester paper “Stickiness” Chris first mentions RI as an information-theoretic explanation for inertia in prices and quantities – one derived from an optimising constraint and not imposed exogenously. About five years later the RI paper was out, and it had a large and lasting impact on the macroeconomics of information frictions.

Chris was extremely generous with us during our PhD years and throughout our academic careers. Many, many others shared the same experience in terms of advice and feedback. To be sure, talking to Chris was not exactly like getting feedback from AI. Praise and compliments were rare; more often than not, after our conversations we felt somewhat dejected, as he had laid bare the flaws and shortcomings of our work, and/or pointed us toward directions that were not easy to grasp for graduate students (or, for that matter, for anyone without his exceptional intellect). But he always did so with patience, respect, and care. Chris cared deeply about his students and about people more generally. Most importantly, his feedback made us better economists and pushed us to produce better work. For the “gravel in our guts”, to quote Johnny Cash, we have him to thank. We will miss Chris dearly.

References

Doan, T, R Litterman and C Sims (1984), “Forecasting and conditional projection using realistic prior distributions”, Econometric Reviews 3(1): 1–100.

Leeper, E M and C A Sims (1994), “Toward a Modern Macroeconomic Model Usable for Policy Analysis”, NBER Working Paper 4761.

Sargent, T J (1981), “Interpreting Economic Time Series”, Journal of Political Economy 89(2).

Sargent, T J and C A Sims (1977), “Business cycle modeling without pretending to have too much a priori economic theory”, Federal Reserve Bank of Minneapolis Working Paper No 55.

Sims, C A (1980), “Macroeconomics and Reality”, Econometrica 48(1): 1–48.

Sims, C A (1994), “A simple model for study of the determination of the price level and the interaction of monetary and fiscal policy”, Economic Theory 4: 381–399.

Sims, C A (1998), “Stickiness”, Carnegie-Rochester Conference Series on Public Policy 49: 317-356.

Sims, C A (2003), “Implications of rational inattention”, Journal of Monetary Economics 50(3): 665-690.

Sims, C A (2007), “Bayesian Methods in Applied Econometrics, or, Why Econometrics Should Always and Everywhere Be Bayesian”.

Sims, C A (2016), “Luncheon Address: Fiscal Policy, Monetary Policy and Central Bank Independence”, Jackson Hole Economic Symposium.