The Standard Narrative about DSGE Models in Central Banks’ Technical Reports

Abstract

Historians of macroeconomics, through the analysis of articles in peer-review journals, pointed out macroeconomists’ propensity to elaborate narratives about the history of their discipline. This article extends the analysis of self-produced narratives to a different genre of literature—namely technical reports on DSGE models published by central banks and other policy-making institutions. This literature adopts a narrative displaying two distinctive characteristics: the emphasis on “consensus” (leading to “better microfoundations”) and on “technical change” (enhancing the “fit” between theory and “facts”). Relying on these two arguments, the narrative told in technical reports conveys a rhetorical argument to legitimize the use of DSGE models in policy institutions.

Keywords:

• DSGE models
• central banks
• history of macroeconomics
• microfoundations
• bayesian econometrics

JEL codes:

• B22

Acknowledgments

I would like to thank Pedro Duarte and an anonymous referee for their helpful comments. My work also benefited from the stimulating discussion during the HISRECO workshop (Luzern, April 2017): I am much indebted to the participants and the organizers. I am also grateful to Jean-Sébastien Lenfant, Annie Cot, Jérôme Lallement, and to the participants to the Alfred O. Hirschman seminar (Université Paris 1), in particular to Pierrick Dechaux, Aurélien Goutsmedt, Erich Pinzon-Fuchs, and Matthieu Renault, for their comments on the earliest version of this work.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 In this paper, policy-making institutions include mostly central banks (since they were the leading producers of DSGE models in the past two decades), but also international organisations (such as the International Monetary Fund, IMF).

2 Most technical reports are not officially endorsed by the institution; they are signed by their authors (i.e. those who have been in charge of the development of the model) with the usual caveat (“views expressed in this report are solely those of the authors etc.”). In most cases, technical reports have not been peer-reviewed, although they usually result from a collective writing effort supervised by a senior officer.

3 Early developments of DSGE models include Cooley (1995), Henin (1995), Goodfriend and King (1997), and Clarida, Gali, and Gertler (1999). The label “DSGE” appeared for the first time in press in Rankin (1998).

4 There are actually many microfoundational programs investigating the relationship between individual behaviour and aggregate phenomena (Hoover 2012). Lucas’s program is one particular example, characterized by the representative agent hypothesis and market clearing. Since alternative programs are not addressed here, I will simply use the word “microfoundations” instead of “Lucasian microfoundations” or “representative agent microfoundations”.

5 As traditionally suggested by Clarida, Gali, and Gertler (1999); for a formal derivation of these equations from the individual maximization problems, see Woodford (1998) or, for a simplified version, Walsh (2003, chap. 5).

6 For an early survey of these issues, see Tovar (2008). It should be noted that indeed most of these developments had started even before the 2008 crisis (as stressed by Christiano, Motto, and Rostagno 2008). However, it seems fair to say that since 2008 the DSGE approach devoted a far greater attention to effectively develop these research areas.

7 This line of work seems for the moment the one that has produced the most significant amount of new research. Early contributions are for instance Christiano, Motto, and Rostagno (2003, 2008), Castelnuovo and Nistico (2010), Iacoviello and Neri (2010), and Boissay, Collard, and Smets (2013).

8 A typical illustration of such a change is the evolution of Blanchard’s views, as analysed by Brancaccio and Saraceno (2017).

9 This is also consistent with the findings by Claveau and Dion (2018), about central banks’ increasing engagement with academic research starting from the end of the 1990s. As Douglas Laxton, an IMF official, put it, “Much of the success” of DSGE models in policy-making institutions “has been a result of their strong links to the academic literature” (Laxton 2008, 214).

10 For a more comprehensive overview of the different uses of DSGE models—policy analysis (“conditional forecasting”) or forecasting (“now” or “real-time” or “short term” forecasting)—see for instance the survey by Hammond (2015).

11 It is beyond the scope of this paper to illustrate how the spread of DSGE models was actually supported by a dense network of modellers, connecting all these policy-making institutions. But cross-fertilization among policy-making institutions did play a crucial role for rise of DSGE models across the world. Modellers building ECB, Fed Board and IMF’s first DSGE models were subsequently involved, directly or indirectly (consultancy, co-authoring), with the construction of DSGE models for other policy-making institutions around the world. An early example is Czech National Bank “New Model” (or “G3”), which was designed relying on the structure of IMF’s GEM model, with the active support of IMF researchers (Beneš et al. 2005).

12 For a case study see Clinton et al. (2017). Note also that some DSGE modellers claim that the DSGE approach was able to adapt to the new policy environment of the 2008 crisis, i.e. assessing and advising non-conventional monetary policy that followed the 2008 crisis (forward guidance, quantitative easing; see for instance Lindé 2018, 274–275 for a review). Christiano, Eichenbaum, and Trabandt (2018, 129–130) claim also that DSGE models can be adapted to assess fiscal policy scenarios within a zero-lower-bound-type of environment.

13 A significant development occurred with respect to the analysis of the financial and banking sector. The ECB has been leading this process, developing and using very early a version of NAWM including a financial block (McAdam and Lombardo, 2009) and developing an alternative DSGE model focused precisely on the financial sector (the “CMR model”, inspired by Christiano, Motto, and Rostagno 2008, 175). See Smets et al. (2010) for a review of these two models.

14 These hearings were organised in 2010 by the Committee on Science and Technology to investigate “the appropriate roles and limitations of models such as DSGE models” (Broun 2010, 1). Many scholars have been already commenting on these hearings (see for instance De Vroey 2016, chap. 20).

15 The ECB embraced again “old” structural macroeconometric models (such as the “multi-country” model by Dieppe, Pandiella, and Willman 2011). Also, the ECB is considering the development of alternative approaches (such as agent-based modelling) in the future, as stated by Vitor Constâncio (ECB vice-president) in his opening speech to the second annual ECB research conference (Constâncio 2017). The Bank of England also emphasizes since 2013 a “multiple models” approach, although the DSGE model COMPASS kept a central place in the toolbox for policy analysis and forecasting (Burgess et al. 2013). A similar call for the use of a “set of models” was made famously by Blanchard (2016), and then by most participants to the recent Oxford Review for Economic Policy symposium.

16 Since the crisis, we should also acknowledge a rather “defensive” attitude, which tend to legitimize DSGE models while recognizing their limitations, and opening the door to some alternative approaches (see for instance Lindé 2018). This new attitude fits with the general atmosphere of a renewed “plurality of models” as described above (fn. 17).

17 https://publications.banque-france.fr/en/liste-chronologique/banque-de-france-bulletin [retrieved on 23/04/2019]

18 A similar use of the adjective “modern” (as opposed to “traditional”) macroeconomics could also be found in textbooks (see for instance Chugh 2015, 170).

19 And some did not: Christiano, Eichenbaum, and Trabandt (2018) famously started with calling criticists of the DSGE approach “dilettants”, then continued with arguing that most criticisms against DSGE models were “not informed criticisms”, and finally concluded by restating that “There is simply no credible alternative [to DSGE models]”. Note that the sentence about “dilettants” disappeared from the published version of the article.

20 This seems the most complete definition of “progress” that can be found in the various self-produced narratives about macroeconomics. It is then a loose definition, almost a commonplace notion; those using it do not seem to have engaged any further with the literature discussing this topic (see for instance Lawson 1987; Backhouse 1997 or Bridel 2005 ).

21 Also note the overabundant use by Blanchard of the word “progress” in his textbooks (Blanchard, Giavazzi, and Amighini 2013; Blanchard and Johnson 2013, resp. chap. 24 and chap. 25). The idea of “progress” is mentioned and discussed eight times in four pages—including the emphatic paragraph headline “Progress in all fronts” referring to the evolution of macroeconomics during the first two decades of the post-war period.

22 Nevertheless, the use of self-produced narrative seems to enjoy a distinctive popularity in macroeconomics with respect to other sub-fields (except maybe for international trade). For instance, if we compare a sample of the most common, recent textbooks on macroeconomics (Heijdra and van der Ploeg 2002; Walsh 2003; Dornbusch, Fischer, and Startz 2007; Wickens 2012; Burda and Wyplosz 2013; Blanchard and Johnson 2013; Blanchard, Giavazzi, and Amighini 2013; Jones 2014; Chugh 2015; Mankiw 2016) with a sample of equally common textbooks on microeconomics (Varian 1992, 2009; Pindyck and Rubinfeld 2008; Ruffin and Gregory 2000; Frank 2006; Mankiw 2017), the comparison outlines the following. We can notice that, of the microeconomic textbooks, none addresses the history of microeconomics. Conversely, all most common macroeconomics textbooks address the history of macroeconomics—although the attention devoted to history is variable (from a whole chapter, as in Blanchard’s textbook, to scattered remarks, as in Burda and Wyplosz 2013; Mankiw 2016). This suggests that historical narratives play a greater role in the teaching of macroeconomics. Although this is not the place for conducting an in-depth analysis of this issue, it could be argued that historical narratives could play a similar rhetorical role in technical reports and in textbooks.

23 See also De Castro et al. (2011, 6) about the “SAMBA” model (Central bank of Brazil).

24 Looking at the past with a retrospective and a teleological standpoint is typical of “rational reconstruction” implied by the “whig history” (Blaug 2001). Hence, past macroeconomic models are presented and assessed using the standards of current DSGE models—hence, past models are described as “primitive” with respect to “modern” models. This is also the role granted to history in macroeconomics textbooks: “We’ll begin by tracing the historical development of [DSGE] models. It’s a great way to understand some of the limitations of the early models and how they have evolved—and continue to evolve—to overcome these limitations.” (Jones 2014, 407)

25 A point also raised in several textbooks, as for instance in Jones (2014, 409): “[RBC theory] led to an explosion of additional research as economists sought to enrich the models to include other shocks and explain other economic variables.”

26 See also Wickens’s textbook (Wickens 2012, xiii): “DSGE macroeconomics has emerged in recent years as the latest step in the development of macroeconomics from its origins in the work of Keynes in the 1930s.”

27 For instance, Woodford peculiarly insists on the role of Wicksell in pioneering the DSGE approach, while most other DSGE modellers (including those writing technical reports) never mention this filiation. Also, authors change their narratives: Lucas, for instance was use to trace back the origins of his approach to Hayek (Lucas 1977), then he abandoned that reference (Lucas in Snowdon and Vane 1998, 121), and, in his Nobel lecture, he granted an important role to David Hume (Lucas 1996).

28 We could assume that an alternative rationale for the standard narrative is to focus on the history of modelling within policy-making institutions—this could for instance explain why Keynes itself is absent from the narrative, while macroeconometric modelling à la Klein and Goldberger (1955) are present. However, this assumption is weak, as most of the steps in the standard narrative (new classical macro, RBC, new Keynesian economics) do not correspond at all to modelling practices adopted in policy-making institutions.

29 Or, as Azariadis and Kaas (2007, 14) put it, DSGE models are the “unifying platform” for macroeconomics, playing “a similar role to [the one played] by string theory in modern physics”. A similar argument can also be found in a recent macroeconomics textbook by Wickens (2012, xv): “The virtue of DSGE macroeconomics is brought out by the following encounter with a frustrated student. He protested that he knew there were many theories of macroeconomics, so why was I teaching him only one? My reply was that this was because only one theory was required to analyse the economy, and it seemed easier to remember one all-embracing theory than a large number of different theories.”

30 As for “progress”, the notions of “schools” and “revolution” are loosely defined in these narrative (see Duarte 2016).

31 Although without using the terminology of “schools”, another similar argument has been made by Robert Hall (1976), with the distinction between “salt-water” and “fresh-water” (or “clear-water”) macroeconomics. This is another topos in textbooks (see for instance Burda and Wyplosz 2013, 16).

32 Of course accumulation of knowledge (“progress”) can also take place through revolution, as claimed, for instance, by Rodano’s in “Contemporary Controversies in Macroeconomics” (Rodano 2002, 307): “the disputes, debates, skirmishes and head-on battles between scholars played a constructive role in the progress of the discipline. […] discussion in macroeconomics, far from being sterile, has actually favoured a real improvement of the discipline”.

33 In other genres, the two rhetoric sometimes even coexist within a same paper, as for instance in Vines and Wills (2018).

34 In a way, the choice of the standard narrative to emphasize consensus can be seen as vindicating Woodford (1999, 2)s claim that progress in macroeconomics “is far from transparent”, as “macroeconomics has been famously controversial”.

35 This also echoes Hall (1976)s much quoted “saltwater” and “freshwater” divide: “The freshwater view holds that […] government is essentially incapable of affecting the level of economic activity. The saltwater views […] thinks government policies (at least monetary policy) is capable of affecting demand.”

36 In the quote, controversies among political agendas: but, in the revolution view, political disagreements among schools of thought are frequently assimilated with their theoretical and methodological disagreements—as it is for instance explicit in Hall (1976)s distinction between salt and fresh water macroeconomics. As also noted by De Vroey and Duarte (2013), the “old” neoclassical synthesis was also rather an agreement about policy than theory.

37 “Quarterly Projection Model”, the model for forecasting previously in use at the Bank of Canada.

38 In Christiano, Eichenbaum, and Trabandt (2018, 123) we also read that “technological constraints were real and binding”, this time with respect to solving and estimating non-linear models.

39 Note, however, that this assessment has been challenged many times. To take a recent example, Hendry and Muellbauer (2018, 303–308) pointed out the relatively weaker data-week performance of the Bank of England BEQM DSGE model compared to the previous “MTMM” model.

40 Pagan (2003) expresses a similar view. Erceg, Guerrieri, and Gust (2005, 1) and Bayoumi (2004, 2) suggest that, during the 1990s, this dilemma originated a divide between academic modelling (oriented by theoretical concerns) and modelling in policy-making institutions (oriented by empirical, i.e. “data-fit” concerns).

41 Surprisingly, VAR modellers hold a very similar perspective: according to them, the evolution of macroeconomics results from a tension between “theory-driven” and “data-driven” models (Spanos 2009; Juselius 2010).

42 Forecasting and Policy System, the model previously in use at RBNZ.

43 Fernandez-Villaverde’s point is that Bayesian econometrics relies on integrating a maximum-likelihood function (instead of maximizing it) and that integration can be more easily computed by a software than maximization.

44 Although the main audience of such a disclosure remained limited to the academic and policy-making sphere, some authors include this element in their account of the DSGE approach as fostering central banks “transparency”, “openess”, and “accountability” toward the general public (see e.g. Clinton et al. 2017; Christiano, Eichenbaum, and Trabandt 2018). Open-source models are also praised since they significantly decrease the cost associated with model’s “housekeeping” (update, revisions, extensions) and with “transferring” the model to other users within the same institution or across institutions (see for instance Lindé 2018, 277–278).