![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
By focusing on the literature on the estimates of the natural rate of interest, this paper critically analyses the modern practice of identifying the benchmark rate of monetary policy with an equilibrium or neutral interest rate reflecting “fundamental forces” unaffected by monetary factors. After briefly mentioning the determinants of the natural rate of interest in the New Keynesian models, the paper discusses the different notions of it that we find in these models and the problems encountered when the natural rate is estimated. It states that these problems are not only related to the difficulties in distinguishing the kind and persistency of economic shocks, but pertain to theory, namely to model specification and the alleged independence of the average or normal interest rate from monetary policy. Following Keynes’s suggestion regarding the monetary nature of interest rates, some final remarks will be advanced on their effects on prices and income distribution as well as on the objectives and stance of monetary policies
Notes
1 According to Wicksell, the bank system tends to behave in a routine, conservative way, leaving the interest rates unchanged when a change in the natural rate occurs, for instance due to technical progress. He thus advanced an “exogenous” explanation of the discrepancy between the market and natural interest rates unlike, for instance, von Mises who argued that credit cycles stem from the tendency of central banks to lower interest rates below the natural ones.
2 Wicksell maintains that this elasticity increases with bank concentration, the issuing of bills and notes and the development of clearing methods. He thought, however, that rising prices would eventually lead to an increase in interest rates because the demand for cash holdings impinges on its supply (see Wicksell Citation1898, 110) and the profit margins of the bank system will otherwise begin to shrink.The (temporary in nature) previous investment decisions will thus be reverted to the normal amount determined by the natural rate of interest, namely by the real interest rate that is expected when choosing the cost-minimising technique and all resources are fully employed with stable prices.
3 Both Pigou and Fisher before Friedman explained trade cycles by referring to money illusion. Thus, according to Fisher (Citation1930, 285), an increase in money supply will lead to an increase in prices and profits, but initially the rate of interest does not increase because the fact that borrowers are able to pay higher nominal interest rates is not understood. The consequent extra-profits will lead to a rise in the demand for loans and therefore the interest rate begins to increase, which also explains, according to Fisher, the Gibson paradox, namely the co-movement of interest and prices (Fisher Citation1930, 282).
4 Typically, households are the owners of firms, representative agents are assumed and there are complete financial markets.
5 The value of k depends on the fraction of firms that adjusts price in any period and from the strategic complementarity between them. Together with the demand elasticity for monopolistic firms and the inverse of Frisch elasticity of labour supply, it shapes the slope of the Phillips Curve.
6 This arises from first order conditions of constrained utility maximisation. With additive intertemporal utility functions, it states that the marginal utility of consumption at time t is equal to (1 + r)/(1+ times its marginal utility at time t + 1, where r is the rate of interest and
is the rate of inter-temporal preference, namely a household's preference to anticipate consumption.
7 More generally, in the DSGE models utilised by the central banks to forecast the effects of monetary policies and the natural rate of interest, the public sector and the “rest of the world” are also taken into account in addition to households, firms and the monetary policy maker (see Bank of England Citation2013; Del Negro, Giannoni, and Schorfheide Citation2015). Therefore, the output gap would also reflect the negative effects on net exports of an appreciation of the real exchange rate driven by an increase in the interest rate.
8 The specific relation shaping the natural rate of interest depends on the assumptions of the model including that on household preferences, their degree of altruism, the influence of habit formation and the weight of liquidity constrained households (for whom current rather than future income is relevant for present consumption). For examples of this class of models, see Christiano et al. (Citation2005), Smets and Wouters (Citation2007), Giammarioli and Valla (Citation2003).
9 As noted by Woodford (Citation2003, 69) “[t]he predictions of the neo-Wicksellian theory [are not] really (….) different from those of a standard quantity-theoretic analysis” when the Taylor principle is introduced to avoid price indeterminacy (see also Bullard and Mitra Citation2002).
10 It is stated that this would conform to Keynes’s reference in the General Theory (1936) to a ‘neutral’ rate of interest that is ‘consistent with full employment, given the other parameters of the system’. However, as will be specified below, if this remark stemmed from some neoclassical elements that were still present in his analysis, Keynes emphasised the monetary nature of the interest rate and did not ascribe unemployment to nominal price rigidities as in the New Keynesian models.
11 This is the reason why central banks should change the rate of interest slowly, for instance by a rule according to which For a critical analysis of the different specifications of the Taylor rule, see Levrero (Citation2019).
12 Blinder specifies that unrepresentative periods should be excluded from the average such as the 1970s when real short-term interest rates became negative or the 1980s when they were exceptionally high. Considering these averages for 30–50 years, we would get a value of between 1.75 and 2.25 per cent as the normal or neutral rate of interest. Blinder judged the stance of US monetary policy on this basis. For instance, it would have been accommodative in the years 1990–91, when the real rate of interest approached zero which “is well below the neutral rate by anyone’s reckoning” whereas it would have been restrictive around 1994–95 and neutral around 1997.
13 A similar concept also appears in the final chapters of Keynes’s Treatise on Money (see Panico Citation1987).
14 It has been a natural development of the NRI as the time invariant intercept of the Taylor rule in an identified structural VAR (see Rotemberg and Woodford Citation1997) where potential output and the NRI are taken as known in order to identify the shocks of monetary policies. According to Rudebusch (Citation2001), the monetary authorities would in fact do their “experiments” using an average rate of interest but facing a variable IS curve in each period.
15 Here the idea is that long-term interest rates embody expectations regarding future short-term rates and that the latter are linked to “equilibrium” rates (see Bernanke and Blinder Citation1992). Therefore, if the yields curve steepens, the short-term interest rate will be lower than the natural rate, whereas if it flattens, the opposite will be true. However, as Bomfin (Citation2001) acknowledges, the long-term interest rates may vary for reasons not linked to changes in the differentials between actual and equilibrium short-rate interest rates, for instance, due to inflation expectations that can lead to higher long-term rates even if the underlying real rates are unchanged. Furthermore, even if inflation-indexed bonds are considered, the problem remains of possible distortions due to liquidity premium, term and risk premiums, irrational expectations, and the fact that only short series exist for these rates. See also Blinder (Citation1998, 71) who stressed that forward rates implicit in the long-term rates are poor forecasters of future rates.
16 Laubach and Williams (Citation2003) and Wu (Citation2005) provide an application of Hodrick-Prescott and band-pass filters to estimate the natural rate of interest whereas Crespo, Cuaresma, Gnan, and Ritzberger-Gruenwald (Citation2004) and Hamilton et al. (Citation2015) use multivariate techniques to decompose trend and cycle.
17 Thus, the increasing inflation rates during the 1960s and 1970s would indicate a rate of interest that is lower than the natural one, whereas during the years of Volcker’s disinflation, the opposite would have occurred. Univariate time series in particular could not control for this and would ascribe these patterns of the interest rate to its trend.
18 Assuming that the vector ξ at time t of the logarithmic deviations of the model variables from their steady state values due to temporary shocks ε is a function of these shocks and the vector ξ at the previous time
where G and M are matrices whose elements depend on the vector of the structural parameters of the model, these parameters and the unobservable variables such as the natural rate of interest are usually estimated by means of a Bayesian approach. It combines prior beliefs (from theory or from information obtained from the estimated sample) on the values of some of the parameters with the maximum likelihood function to form posterior densities from which the values of the parameters
are picked up. With the Kalman filter, the unobservable variables are then estimated updating progressively (see Harvey Citation1989) their one-sided values within the time horizon of the sample.
19 The augmented specification of the model may also imply distinguishing between residential and non-residential investment, liquidity constrained and unconstrained households, durables and non-durable goods, and so on.
20 Note that this volatility contrasts with the prudence of central banks in changing the rate of interest due to the awareness of the effect of this change on the stability of financial markets.
21 With this estimate method, the NRI estimates are in fact influenced by prior beliefs regarding the deep parameters that shape, for instance, the slope of the IS and Phillips curves, such as the intertemporal elasticity of substitution, the household discounting rate and the wage elasticity of labour supply.
22 This does not exclude using a Bayesian method, with prior information on some variables and parameters, for instance on potential output as derived by a production function as in Pescatore & Turunen (Citation2015).
23 For the meaning of the symbols, see EquationEquations (1–4). In Laubach & Williams (Citation2003) the semi-structural model is similar. In the NRI EquationEquation [8](8)
(8) , the term c appears and the Phillips curve includes two variables concerning the shocks on imported and fuel prices. Moreover, to better identify output fluctuations, a relation between private work hours and the output gap is introduced.
24 More generally, the state-space model is
where y and x are the endogenous and exogenous variables respectively, ξ is the state vector of the unobservable variables, v and ε are Gaussian and mutually non-correlated errors with mean zero and matrices Q and R of variances and covariances. Finally, matrices A and H have as their elements the coefficients of current and lagged exogenous and state variables.
25 More precisely, its reciprocal c as defined in EquationEquation (3)(3)
(3) is equal to one. It is in fact acknowledged that the relationship between the NRI and the growth rate is not well identified in the data.
26 As known, the Kalman filter, is a recursive algorithm that uses a series of measurements observed over time and produces estimates of unknown variables. In the prediction step, the Kalman filter produces estimates of the current state variables, along with their uncertainties. Once the outcome of the next measurement is observed, these estimates are updated using a weighted average, with more weight being given to estimates with higher certainty. The algorithm is run in real time, using only the present input measurements and the previously calculated state and its uncertainty matrix. See Harvey (Citation1989) and below, page 18.
27 The greater these values, the larger the uncertainty on the trend of g and the unobserved process z.
28 In particular, the variability of the NRI increases if the z process is not a random walk. With this hypothesis, the NRI would increase just after the 2008 crisis and be at 1.8 percent in 2016 in contrast to the non-recovery path estimated by Laubach and Williams.
29 Holston, Laubach and Williams (2016) view these “as minimal priors on the structure of the model that, in the event, facilitate the convergence of the numerical optimisation during estimation.” They dictate that must be higher than or equal to 0.025, and
be lower than or at the most equal to – 0.0025. On the other hand, a flat IS curve would increase uncertainty over the value of the NRI (see ECB Citation2018, 23) whereas an insensitivity of output to the interest rate would impair the estimate process of the NRI, as argued in the next Section.
30 Also, real long-term interest rates fell. The average real interest rates on ten-year government bonds for the G7 countries (excluding Italy) passed from 3 to 4% in the years 1987–1999, to 1–3% in 2000s, to 1–0.5% in the years 2000–2015, with even negative values in recent years (see King and Low Citation2004). The liquidity and risk premiums on these bonds dropped due to quantitative easing and forex interventions facing the fall in the net supply of safe assets during the recent financial crises (see again King and Law Citation2004; Caballero and Fahri Citation2014).
31 It is usually stated that the NRI was higher than the actual rates in the 1960s and 1970s and lower in the 1980s and 1990s. However, Orphanides (Citation2001) contested Taylor’s idea that before Volcker, the Fed did not follow the principle of increasing the rate of interest more than inflation, thus determining macroeconomic instability.
32 Another difficulty stems from the continuous revisions of the estimates which in the years 2008–2013 led to a progressive lengthening of the adjustment phase of the short natural rate to its long or steady state value.
33 This refers to the Orphanides and Williams (Citation2002) rule
34 On this possibility see, for example, Taylor (Citation1993), Blinder and Reiss (Citation2005), Hetzel (Citation2015). It has often led to adjusting the values of the parameters and the unobservable variables in order to fit better the data, resulting in a variety of possible monetary policy rules with different policy implications and a loss of their prescriptive role.
35 According to Taylor and Weiland, Laubach’s estimates would indeed underrate the value of m and Fed’s deviation from Taylor rule and the NRI should be left unchanged at its steady state level of 2 per cent.
36 Summers (Citation2014) states that in the US, the convergence of actual to potential output in recent years also occurred thanks to an estimated fall in potential output, so that, whereas in 2014, the actual output was 10 per cent lower than the value predicted in 2007 for this year, potential output fell by 5 per cent. In the meantime, even if the unemployment rate moved towards 5 per cent, the employment to population ratio remained below the pre-crisis level. The fall of potential output is seen to stem directly from the fall in capital accumulation and indirectly from its negative effects on total factor productivity and participation rates. See also Ball (Citation2014), Reifschneider, Wascher, and Wilcox (Citation2013), Stockhammer and Sturns (Citation2011).
37 Note that in the New Keynesian models, supply factors mainly affect the NRI with the exception of shocks on time-preference shifting the saving function which are usually interpreted as demand shocks. The possibility that effective demand has an influence on productive capacity and labour productivity opens an identification problem regarding the nature and effects of frictions and shocks. As stressed by Reifschneider, Wascher, and Wilcox (Citation2013, 26), the statistical methods used to distinguish between cycle and trend hide these problems.
38 However, by itself the payment of a lower amount of interests on public debt may reduce total expenditure according to the propensities to consume of those who receive these interest payments.
39 What matters for residential and fixed capital investments is, of course, the long-run interest rate and, as noted by Kaldor (Citation1982), inventories are often financed by long-term capital too. The relation between the (nominal) short-run policy rate and the (real) long-run interest rate is thus crucial for the transmission of monetary policy. If they diverge for short periods, central banks have, however, instruments to influence the long run interest rates by shaping expectations on the future short rates, modifying net supplies of safe and risky assets through open market operations, varying the assets admitted as collateral and their haircuts when financing the bank system, thus affecting liquidity and risk premia. See, in this respect, Deleidi and Levrero (Citation2020a).
40 The lack of sensitivity of consumption to the interest rate is nowadays also stressed in models with microeconomic heterogeneity (Kaplan and Violante Citation2018).
41 See the brief account of the results of the capital controversy below.
42 This is the prevailing interpretation of the recent experience of negative real policy interest rates and income growth rates that are only slightly higher than its estimated trend. As noted by Lubick and Matthes (Citation2015), distorted estimates of the NRI may arise also from misspecification of the Phillips Curve which may also become negative.
43 For a criticism of the idea (for example, Sims Citation1992, Castelnuovo and Surico Citation2010) that, if the price puzzle emerges, it is only due to specific monetary policy regimes and/or information omitted on the systematic part of monetary policy in the estimated VAR models, see Barth and Ramey (Citation2001), Christiano, Eichenbaum, and Evans (Citation2005) and Deleidi and Levrero (Citation2020).
44 We do not consider here the possibility of forced saving or a destruction of productive capacity stemming from interest rates that are different from the NRI. In the neo-Wicksellian literature, these possible channels of adjustment of the NRI to the market interest rates rather than the opposite are usually neglected or have a temporary nature.
45 Note that Woodford (Citation2003, 167 and 353) maintains that, in the short run, capital goods should be treated as specific inputs with different real rental rates. Sometime he also refers to an economy with a single good (see idem, 41). However, the reference to capital as a value magnitude is implicit when he argues that there is a tendency of capital to shift towards sectors with higher capital returns and that investment goods are perfect substitutes for savers (see Woodford Citation2003, 166). It is also implicit in the investment functions with adjustment costs of the New Keynesian models or when it is argued that firms borrow at a rate equal to the value of the marginal product of capital.
46 As the Sonneschein-Mantel-Debreu theorem shows, also in the neo-Walrasian versions of the general equilibrium theory, where prices are distinguished by date of delivery, the existence, but not unicity and stability, of the equilibrium is proven, unless restrictive assumptions are introduced (see Kirman Citation1989). Moreover, price indeterminacy may arise in a sequential economy for a class of linear models when the number of consumption goods is lower than the number of inputs (see Mandler Citation1995, and Fratini and Levrero Citation2011). While challenging the significance of the theory (see Petri Citation2004), the neo-Walrasian abandonment of the notion of long-run equilibrium would thus not solve the issues mentioned above concerning the implementation of monetary policies.
47 This point is different from the possibility of global indeterminacy of the price level outlined by Benhabib, Schmitt-Grohe, and Uribe (Citation2001) which would imply that monetary authorities ought to influence price expectations in order to achieve local determinacy.
48 While wage bargaining acts on money wages, monetary policies fix the nominal interest rates. If the latter remain unchanged, a continuous increase in money wages will lead to an increase in real wages which will be higher the lower the share of wages in the gross product. This does not happen if the monetary authorities react by increasing the nominal rate of interest in order to maintain a desired real interest rate (see Levrero Citation2013). The consequent accelerating inflation will come to an end whenever this rate is lowered by the monetary authorities or increasing prices and nominal interest rates lower workers’ bargaining power by raising the rate of unemployment or reducing their degree of organisation. It may also end if a rise in labour productivity or improvements in the terms of trade allow workers’ wage requests to be reconciled with the monetary policy benchmark rate.
49 As in after the collapse of the tech bubble at the beginning of 2000, or after the 2008 crisis, low interest rates are often implemented to sustain consumption and thus capital accumulation. As in after the Second World War (see Kaldor Citation1982), this “low gear regime” usually leads to increasing pressure from the bank sector and rentiers for a rise in interest rates, especially whenever it has been associated with lowering profit margins of the bank sector. In these circumstances, central banks act to guarantee the reproducibility of the economy over time and do not necessarily pander to the requests of the financial sector, while assuring it the liquidity it needs.
50 The uncertainty also stems from the fact that prima facie an increase in the interest rate leads to an increase in prices. A lower rate of inflation depends thus on its effect on the course of money wages.
51 Different normative proposals have been advanced in this regard from the Kansas City rule of an interest rate leading to the euthanasia of the rentier (Wray Citation2007; Forstater and Mosler Citation2004), to the Smithin rule (Citation2004, 686) of a real rate of around 2 per cent. Sawyer advances the proposal of a “fiscal real rate,” namely, a rate that is lower than the growth rate of the economy in order to leave space for expansionary fiscal policies without increasing public debt, whereas Lavoie and Seccareccia (Citation1999) propose a nominal “fair rate” equal to labour productivity growth rate plus the inflation rate. For a discussion of these different proposals, see Levrero (Citation2020), Lavoie and Seccareccia (Citation2020) and Pivetti (Citation2020).
Additional information
Notes on contributors
Enrico Sergio Levrero
Enrico Sergio Levrero is Associate Professor of Economics at Roma Tre University, Rome, Italy, and editor of the Bulletin of Political Economy. A preliminary version of this article appeared in the INET Working Paper Series. He would like to thank the Institute for New Economic Thinking (INET) for a grant supporting this research project. He would also like to thank Roberto Ciccone, Mario Seccareccia and two anonymous referees for their useful comments. The usual caveat applies.