This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
Other variants of Keynesian models assume instead of sticky wages that prices are sticky. See Romer (1996: 214ff.), for an extensive discussion.
See Romer (1993: 5) on these two points.
This section draws on the discussion of the Phillips curve in Espinosa-Vega and Russell (1997: 6ff.).
This figure is reproduced from Espinosa-Vega and Russell (1997: 7).
For a more detailed discussion of the Phillips curve, see Espinosa-Vega and Russell (1997: 6ff.).
See Goodfriend and King (1997: 236), for a discussion of the empirical evidence on the Phillips curve in the 1950s and 1960s.
Fisher and Seater (1993: 402) define long-run neutrality (LRN) and long-run super-neutrality (LRSN) of money as follows: “By LRN, we mean the proposition that permanent, exogenous changes to the level of the money supply ultimately leave the level of real variables and the nominal interest rate unchanged but ultimately lead to equiproportionate changes in the level of prices and other nominal variables; by LRSN, we mean the proposition that permanent, exogenous changes to the growth rate of the money supply ultimately lead to equal changes in the nominal interest rate and leave the level of real variables unchanged.”
For a detailed discussion of the permanent output-inflation trade-off see Romer (1996: 222ff).
This section draws on De Long and Summers (1988: 437ff). See also Tobin (1996: 4ff).
For a Keynesian approach to measuring the output gap see also the peak-to-peak method in De Long and Summers (1988: 457ff.).
An interesting discussion of the difficulties of demand management policies in reconciling the expectations of firms and trade unions is found in Sachverstandigenrat (1975: 6).
The following line of argument draws on Espinosa-Vega and Russell (1997: 8ff.).
However, Tobin (1993) points out that this behavior would be rational if workers did not care so much about their absolute wage but more about their wage relative to their co-workers. Thus, a worker might be unwilling to accept a nominal wage cut since he does not know for sure if his co-workers will do the same. An increase in inflation, in contrast, ensures that the real wages of all workers are affected in essentially the same way.
The following section draws on McCallum (1989: 181ff.). Small letters denote logarithms throughout the paper.
See Taylor (2001: 125) for Friedman’s position on expectation formation.
This section draws on McCallum (1989: 182ff.).
For a discussion of the natural rate hypothesis see also Romer (1996: 225ff.). Regarding the role of superneutrality for the monetarist framework, see Espinosa-Vega (1998: 16).
See Goodfriend and King (1997: 238).
See also the discussion in Espinosa-Vega (1998: 16).
The seminal work demonstrating the power of monetary policy is Friedman and Schwartz (1963). Regarding the importance of monetary policy relative to fiscal policy in the monetarist framework see Goodfriend and King (1997: 239; De Long 2000: 91).
See also the discussion of the classical and the Keynesian Phillips curve in Sargent and Söderström (2000: 41) and the discussion in King and Watson (1994: l0ff.).
For a discussion of the role of the expectations-augmented Phillips curve in New Keynesian models see Roberts (1995).
For a discussion of the role of the real balance effect in neoclassical theories see Jarchow (1998: 180ff.). For a discussion of the Keynesian skepticism of the real balance effect see Tobin (1993: 59ff). This issue is also discussed in detail in Chapter 4.
See De Long and Summers (1988: 438). Note that the relation given by (2.7) is closely related to the expectations-augmented Phillips curve given by (2.4b). The only differences are that in (2.7) the supply shock is omitted and the deviation of unemployment from the natural rate is replaced with the deviation of output from its natural level, yt−y.
See the discussion of different approaches towards a supply-side explanation of the increase in unemployment in Europe in Bean (1994: 587ff.). A concise theoretical analysis of the role of these supply side factors for high unemployment in Europe is also contained in Sachs (1986). Siebert (1998) discusses supply side factors contributing in Germany to high unemployment. See also Paqué (1999) for a discussion of causes of structural unemployment in Germany.
See also the discussion in Romer (1996: 226).
Figure 2.4 is reproduced from Espinosa-Vega and Russell (1997: 12, Chart 3).
Stiglitz (1997: 3), for example, uses the term natural rate as synonym for NAIRU.
See the discussion in Espinosa-Vega and Russell (1997: 11ff.).
This section draws on Gordon (1997). For a review of the “history” of the triangle model, see Gordon (1997: 18ff.). This model has recently been employed, for example, by the OECD to estimate the NAIRU for several OECD countries. See OECD(2000: 155ff.).
See Gordon (1997: 14). The lag polynomial a(L), for example, denotes a(L) = a0 + a1L + a2L2 +... + anLn.
For a model with a wage variable as an additional determining variable, see Franz (2000: 3ff.).
See the discussion in Gordon (1997: 16ff.).
The following discussion draws on Franz (2000: 5ff).
See also the discussion of these NAIRU concepts in the report of the OECD (2000: 157). The OECD calls the no-shock NAIRU the long-term equilibrium unemployment rate and the NAIRU which is consistent with stabilizing inflation at its current level the short-term NAIRU.
See Franz (2000: 6ff.) for an extensive discussion of this issue.
See Elmeskov (1993: 94), Elmeskov and MacFarlan (1993: 85), and the discussion of his method in Fabiani and Mestre (2000: 14ff).
See also the discussion in Fabiani and Mestre (2000: 15).
See for example OECD (2000) and Fabiani and Mestre (2000).
For a recent application of the NAIRU concept to German data see Franz (2000).
The seminal paper in this regard is Staiger et al. (1996).
The recession dates are taken from Artis et al. (1997), who developed a procedure to determine peaks and troughs in the business cycle similar to the NBER classification procedure for the United States. They propose classical business cycle turning points for the G7 and a number of European countries based on time series of industrial production for the respective countries. A recession is defined as the time period between a peak and the following trough. For Germany, the authors determine the business cycle turning points for the time period beginning in 1961 and ending in 1993. Döpke (1999) uses their procedure to determine the turning points in Germany for the time period from 1994 until 1999. I am grateful to Jörg Dopke for making his results available to me.
Here we apply the band-pass filter to the monthly annualized rate of change of the consumer price index defined as Δpt = 1,200ln(Pt/Pt−1). For a similar investigation for the USA see King et al. (1995). To account for the start-point and end-point problems of these filter methods, we drop the first three and the three last years of the sample period. See also the discussion in Baxter and King (1995: 9) of this issue.
Regarding the second sample period, it is a striking finding that the trend components of the unemployment rate and the inflation rate are extremely highly correlated. Moreover, the negative coefficient is in accordance with the predictions of the traditional Phillips curve. This finding differs markedly from results for the United States. King et al. (1995) have applied the same technique to U.S. data and find no correlation between the trend components in the time period from 1974 until 1992, while the corresponding correlation coefficient for Germany for this time period is −0.83. For the correlation of the cyclical components King et al. report a correlation coefficient of approximately −0.60 over all sample periods, which is very similar to our results for Germany, indicating that Germany and the United States differ mainly in their long-run response to demand shocks.
The following discussion is based on Bullard (1999: 57ff.) and on Fisher and Seater (1993: 402)
For a more detailed discussion see Fisher and Seater (1993: 405ff.).
Fisher and Seater (1993) show that the relative order of integration in our case implies that long-run neutrality holds by definition.
To illustrate the intuition behind this result, we consider following relationship between the output variable yt and the money variable mt, namely yt = b1yt−1 + b2yt−2 +a0mt + a1mt−1 + a2mt−2 + et, and investigate the implications of a cointegration relationship for the long-run neutrality proposition. We include two lags for each variable to allow for some dynamics. Hansen (1993: 142) shows that in error-correction parameterization this equation becomes Δyt = αt(yt−1 − βmt−1) − b2Δyt−1 + a0Δmt − a2Δmt−1 + et, where α = b1+b2−1) and β = (α0 + a1+a2)/(1 − b1 − b2). The parameter β gives the long-run response of output to an innovation in the money stock, provided there is a long-run/cointegration relationship between the two variables. The existence of such cointegration relationship depends on the loading parameter α. Kremers, Ericsson and Dolado have shown that one can test for cointegration between yt and mt by testing the significance of a (see the discussion in Hansen 1993: 148). If α is significantly larger than zero, then output responds to a disequilibrium in the money-output relationship. That is, a permanent change in the money stock would lead in this case to a permanent response of output to restore the long-run money-output relationship. This implies that the long-run neutrality proposition does not hold. Hence, evidence for a cointegration relationship between output and money is sufficient to reject this proposition. Regarding superneutrality, one could investigate cointegration between unemployment ut and inflation Δpt by testing the loading parameter α in the error-correction model Δut =α(ut−1 − βΔpt−1) − b2Δut−1 + a0Δ2pt − a2Δ2pt−1 + et. If α turns out to be significantly larger than zero, then a permanent change in the inflation rate would lead to a permanent change in the unemployment rate, implying that superneutrality would not hold.
See also King and Watson (1992) and King and Watson (1997).
This section is based on King and Watson (1994: 13ff.).
King and Watson (1994: 15ff.) show that this result carries over to richer models than the one considered here.
See King and Watson (1994: 15ff.) for a detailed discussion.
This section is based on King and Watson (1992: 7ff.).
See King and Watson (1997: 73). For the conditions which need to hold for the model to be invertible, see King and Watson (1997: 75ff.).
Note that long-run superneutrality cannot be tested within this model because the money stock is, according to (2.27), integrated of order one and not of order two, as is required for superneutrality tests. For a modification of this model allowing for superneutrality tests see King and Watson (1992: 10).
For a survey on identifying restrictions used in the literature, see King and Watson (1997: 76ff.).
The seminal paper in this regard is Blanchard and Quah (1989). For a survey on bivariate SVAR models using long-run neutrality restrictions, see Gottschalk and Van Zandweghe (2001).
See Fisher and Seater (1993: 408). King and Watson (1997: 77) propose the alternative restriction γmy = 1, which would be consistent with a policy aiming at price level stability under the assumption of stable velocity.
See also the discussion in Dolado et al. (1997: 12).
This section is based on King and Watson (1994: 1 Iff.).
Dolado et al. (1997: 8ff.) show that (2.39a), which is interpreted here as representing the Phillips curve, can be derived from a wage-and price-setting model, assuming imperfect competition and a hysteretic mechanism. Furthermore, they show that (2.39b) can be interpreted as an aggregate demand equation.
Since in the Keynesian version of the Phillips curve, which was the starting point of the investigation in Gordon (1970) and Solow (1970), unemployment is an indicator of aggregate demand, both approaches to estimating the long-run slope of the Phillips curve are closely related, but in our model we are more explicit about the identification of the demand shock. Furthermore, we consider the reciprocal of the Phillips curve slope coefficient estimated by Solow (1970).
See also the discussion in King and Watson (1997: 93).
See King and Watson (1994: 18). These authors also note that, following the “price equation” estimation strategy used by Gordon (1970) and other researchers in the Keynesian tradition, equation (39a) can equivalently be estimated by OLS using the reverse regression of Δ2pt onto Δut and relevant lags.
I am grateful to Mark W. Watson for making available his RATS programs used in the King and Watson (1994) paper. These programs are available from his homepage.
See also King and Watson (1997: 90ff.) for an application of this methodology as superneutrality tests for U.S. data.
To obtain a better estimate of the long-run trade-off it would appear promising to augment the Phillips curve models with other (exogenous) variables to control for shocks to the inflation variable which are not related to the Phillips curve model. For such an extension, see Dolado et al. (1997) or King and Watson (1994: 27ff.). Moreover, Weber (1994: 20) finds strong evidence against a vertical Phillips curve in a Keynesian Phillips curve model. His results are discussed in more detail below.
See also the discussion in Bullard and Keating (1995: 478).
A similar approach to identify the monetarist Phillips curve has been used by Bullard and Keating (1995) and Dolado et al. (1997).
For the sample period 1954–1992 they report a long-run trade-off of −0.29. For the subsample period 1954–1969 they estimate a long-run trade-off of −0.47 and for the period 1970–1992 the corresponding value is −0.23. For Germany, we find for the period 1970–1998 a long-run trade-off of −0.44 for the monetarist Phillips curve based on the “inflation as a monetary phenomenon” identification. See also King and Watson (1997: 95).
Gottschalk and Van Zandweghe (2001) investigate for German data whether the small size of bivariate models adversely affects the reliability of these models. These authors find that the small size is indeed somewhat a problem, but more so for the identification of demand disturbances than for supply disturbances.
For a recent example of a common trend analysis for Germany, see Carstensen and Gottschalk(2001).
See, e.g., Fackler and McMillin (1997) for a detailed description of the historical decomposition technique.
Rights and permissions
Copyright information
© 2005 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
(2005). Keynesian and Monetarist Views on the German Unemployment Problem. In: Monetary Policy and the German Unemployment Problem in Macroeconomic Models. Kieler Studien - Kiel Studies, vol 334. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-37679-8_2
Download citation
DOI: https://doi.org/10.1007/3-540-37679-8_2
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-25650-2
Online ISBN: 978-3-540-37679-8
eBook Packages: Business and EconomicsEconomics and Finance (R0)