THE ECONOMETRICS OF MACROECONOMIC MODELLING

# The supply-side and wage — and price-setting (Chapters 3—8)

In the course of the 1980s and 1990s the supply-side of macroeconometric models received increased attention, correcting the earlier overemphasis on the demand-side of the economy. Although there are many facets of the supply-side, for example, price-setting, labour demand, and investment in fixed capital and R&D, the main theoretical and methodological developments and controversies have focused on wage - and price-setting.

Arguably, the most important conceptual development in this area has been the Phillips curve—the relationship between the rate of change in money wages and the rate of unemployment (Phillips 1958)—and the ‘natural rate of unemployment’ hypothesis (Phelps 1967 and Friedman 1968). Heuristically, the natural rate hypothesis says that there is only one unemployment rate that can be reconciled with nominal stability of the economy (constant rates of wage and price inflation). Moreover, the natural rate equilibrium is asymptotically stable. Thus the natural rate hypothesis contradicted the demand-driven macroeconometric models of its day, which implied that the rate of unemployment could be kept at any (low) level by means of fiscal policy. A step towards reconciliation of the conflicting views was made with the discovery that a constant (‘structural’) natural rate is not necessarily inconsistent with a demand driven (‘Keynesian’) model. The trick was to introduce an ‘expectations augmented’ Phillips curve relationship into an IS-LM type model. The modified model left considerable scope for fiscal policy in the short run, but due to the Phillips curve, a long-term natural rate property was implied (see, for example, Calmfors 1977).

However, a weak point of the synthesis between the natural rate and the Keynesian models was that the supply-side equilibrating mechanisms were left unspecified and open to interpretation. Thus, new questions came to the forefront, like: How constant is the natural rate? Is the concept inextricably linked to the assumption of perfect competition, or is it robust to more realistic assumptions about market forms and firm behaviour, such as monopolistic competition? And what is the impact of bargaining between trade unions and confederations over wages and work conditions, which in some countries has given rise to a high degree of centralisation and coordination in wage-setting? Consequently, academic economists have discussed the theoretical foundations and investigated the logical, theoretical, and empirical status of the natural rate hypothesis, as for example in the contributions of Layard et al. (1991, 1994), Cross (1988, 1995), Staiger et al. (1997), and Fair (2000).

In the current literature, the term ‘Non-Accelerating Inflation Rate of Unemployment’, or NAIRU, is used as a synonym to the ‘natural rate of unemployment’. Historically, the need for a new term, that is, NAIRU, arose because the macroeconomic rhetoric of the natural rate suggested inevitability, which is something of a straitjacket since the long-run rate of unemployment is almost certainly conditioned by socioeconomic factors, policy and institutions (see for example, Layard et al. 1991 ch. 1.3).[2] The acronym NAIRU itself is something of a misnomer since, taken literally, it implies p < 0 where p is the log of the price level and p is the third derivative with respect to time. However, as a synonym for the natural rate it implies p = 0, which would be constant rate of inflation rate of unemployment (CIRU). We follow established practice and use the natural rate—NAIRU—terminology in the following.

There is little doubt that the natural rate counts as one of the most influential conceptual developments in the history of macroeconomics. Governments and international organisations customarily refer to NAIRU calculations in their discussions of employment and inflation prospects,[3] and the existence of a NAIRU consistent with a vertical long-run Phillips curve is a main element in the rhetoric of modern monetary policy (see for example, King 1998).

The 1980s saw a marked change in the consensus view on the model suitable for deriving NAIRU measures. There was a shift away from a Phillips curve framework that allowed estimation of a natural rate NAIRU from a single equation for the rate of change of wages (or prices). The modern approach combined a negative relationship between the level of the real wage and the rate of unemployment, dubbed the wage curve by Blanchflower and Oswald (1994), with an equation representing firms’ price-setting. The wage curve, originally pioneered by Sargan (1964), is consistent with a wide range of economic theories (see Blanchard and Katz 1997), but its original impact among European economists was due to the explicit treatment of union behaviour and imperfectly competitive product markets, pioneered by Layard and Nickell (1986). In the same decade, time-series econometrics constituted itself as a separate branch of econometrics, with its own methodological issues, controversies and solutions, as explained in Chapter 2.

It is interesting to note how early new econometric methodologies were applied to wage-price modelling, for example, equilibrium-correction modelling, the Lucas critique, cointegration, and dynamic modelling. Thus, wage formation became an area where economic theory and econometric methodology intermingled fruitfully. In this chapter, we draw on these developments when we discuss how the different theoretical models of wage formation and price-setting can be estimated and evaluated empirically.

The move from the Phillips curve to a wage curve in the 1980s was, however, mainly a European phenomenon. The Phillips curve held its ground well in the United States (see Fuhrer 1995, Gordon 1997, and Blanchard and Katz 1999). But also in Europe the case has been reopened. For example, Manning (1993)

showed that a Phillips curve specification was consistent with union wagesetting, and that the Layard-Nickell wage equation was not identifiable. In academia, the Phillips curve has been revived and plays a prolific role in New Keynesian macroeconomics and in the modern theory of monetary policy (see Svensson 2000). The defining characteristics of the New Keynesian Phillips curve (NPC) are strict microeconomic foundations together with rational expectations of ‘forward’ variables (see Clarida et al. 1999, Gall and Gertler 1999, and Gall' et al. 2001).

There is a long list of issues connected to the idea of a supply-side determined NAIRU, for example, the existence and estimation of such an entity, and its eventual correspondence to a steady-state solution of a larger system explaining wages, prices as well as real output and labour demand and supply. However, at an operational level, the NAIRU concept is model dependent. Thus, the NAIRU issues cannot be seen as separated from the wider question of choosing a framework for modelling wage, price, and unemployment dynamics in open economies. In the following chapters we therefore give an appraisal of what we see as the most important macroeconomic models in this area. We cover more than 40 years of theoretical development, starting with the Norwegian (aka Scandinavian) model of inflation of the early 1960s, followed by the Phillips curve models of the 1970s and ending up with the modern incomplete competition model and the NPC.

In reviewing the sequence of models, we find examples of newer theories that generalise on the older models that they supplant, as one would hope in any field of knowledge. However, just as often new theories seem to arise and become fashionable because they, by way of specialisation, provide a clear answer on issues that older theories were vague on. The underlying process at work here may be that as society evolves, new issues enter the agenda of politicians and their economic advisers. For example, the Norwegian model of inflation, though rich in insight about how the rate of inflation can be stabilised (i. e. p = 0), does not count the adjustment of the rate of unemployment to its natural rate as even a necessary requirement for p = 0. Clearly, this view is conditioned by a socioeconomic situation in which ‘full employment’ with moderate inflation was seen as attainable and almost a ‘natural’ situation. In comparison, both the Phelps/Friedman Phillips curve model of the natural rate, and the Layard- Nickell NAIRU model specialise their answers to the same question, and take for granted that it is necessary for p = 0 that unemployment equals a natural rate or NAIRU which is entirely determined by long-run supply factors.

Just as the Scandinavian model’s vagueness about the equilibrating role of unemployment must be understood in a historical context, it is quite possible that the natural rate thesis is a product of socioeconomic developments. However, while relativism is an interesting way of understanding the origin and scope of macroeconomic theories, we do not share Dasgupta’s (1985) extreme relativistic stance, that is, that successive theories belong to different epochs, each defined by their answers to a new set of issues, and that one cannot speak of progress in economics. On the contrary, our position is that the older

models of wage-price inflation and unemployment often represent insights that remain of interest today.

Chapter 3 starts with a reconstruction of the Norwegian model of inflation, in terms of modern econometric concepts of cointegration and causality. Today this model, which stems back to the 1960s, is little known outside Norway. Yet, in its reconstructed forms, it is almost a time traveller, and in many respects resembles the modern theory of wage formation with unions and price-setting firms. In its time, the Norwegian model of inflation was viewed as a contender to the Phillips curve, and in retrospect it is easy to see that the Phillips curve won. However, the Phillips curve and the Norwegian model are in fact not mutually exclusive. A conventional open economy version of the Phillips curve can be incorporated into the Norwegian model, and in Chapter 4 we approach the Phillips curve from that perspective. However, the bulk of the chapter concerns issues which are quite independent of the connection between the Phillips curve and the Norwegian model of inflation. As perhaps the ultimate example of a consensus model in economics, the Phillips curve also became a focal point for developments in both economic theory and in econometrics. In particular we focus on the development of the natural rate doctrine, and on econometric advances and controversies related to the stability of the Phillips curve (the origin of the Lucas critique).

In Chapter 6 we present a unifying framework for all of the three main models, the Norwegian model, the Phillips curve and the Layard-Nickell wage curve model. In that chapter, we also discuss at some length the NAIRU doctrine: is it a straitjacket for macroeconomic modelling, or an essential ingredient? Is it a truism, or can it be tested? What can be put in its place if it is rejected? We give answers to all these questions, and the thrust of the argument represents an intellectual rationale for macroeconometric modelling of larger systems of equations.

An important underlying assumption of Chapters 3-6 is that inflation and unemployment follow causal or future-independent processes (see Brockwell and Davies 1991 ch. 3), meaning that the roots of the characteristic polynomials of the difference equations are inside the unit circle. This means that all the different economic models can be represented within the framework of linear difference equations with constant parameters. Thus the econometric framework is the vector autoregressive model (VAR), and identifies systems of equations that encompass the VAR (see Hendry and Mizon 1993, Bardsen and Fisher 1999). Non-stationarity is assumed to be of a kind that can be modelled away by differencing, by establishing cointegrating relationships, or by inclusion of deterministic dummy variables in the non-homogeneous part of the difference equations.

In Chapter 7, we discuss the NPC of Gall and Gertler (1999), where the stationary solution for the rate of inflation involves leads (rather than lags) of the non-modelled variables. However, non-causal stationary solutions could also exist for the ‘older’ price-wage models in Chapters 3-6 if they are specified with ‘forward looking’ variables (see Wren-Lewis and Moghadam 1994). Thus, the discussion of testing issues related to forward vs. backward looking models in Chapter 7 is relevant for a wider class of forward-looking models, not just the NPC.

Econometric analysis of wage, price, and unemployment data serve to substantiate the discussion in this part of the book. An annual data set for Norway is used throughout Chapters 4-6 to illustrate the application of three main models (Phillips curve, wage curve, and wage price dynamics) to a common data set. But frequently we also present analysis of data from the other Nordic countries, as well as of quarterly data from the United Kingdom, the Euro area, and Norway.