The Best of Rules and Discretion: a case for Nominal gdp targeting in India

Download 133.05 Kb.
Date conversion25.05.2016
Size133.05 Kb.
  1   2   3
The Best of Rules and Discretion:
A Case for Nominal GDP Targeting in India

Pranjul Bhandari and Jeffrey Frankel,

Harvard Kennedy School

Aug.6 + Sept.29, 2014


The recent revival of interest in nominal GDP (NGDP) targeting has come in the context of large advanced economies. We argue that the case for NGDP targeting is even more appealing for mid-sized developing countries, because they tend to be more susceptible to supply shocks and terms of trade shocks. For India, in particular, one major exogenous supply shock is the monsoon rains. NGDP targeting splits the impact of supply shocks automatically between inflation and real GDP growth. In the case of inflation targeting (IT), by contrast, the full impact of an adverse supply shock or terms of trade shock is felt as a loss in real GDP alone. NGDP targeting arguably achieves the best of both worlds: it automatically accommodates supply shocks as most central banks with discretion would do anyway, while retaining the advantage of anchoring expectations as rules are designed to do. We outline a simple theoretical model and derive the conditions under which an NGDP targeting regime would dominate other regimes such as IT for achieving objectives of output and price stability. We go on to estimate for the case of India the main parameters needed to ascertain whether these conditions hold, most notably the slope of the aggregate supply curve. We find that under certain plausible conditions, nominal GDP targeting is indeed better placed than IT, especially in the face of the supply shocks that developing countries tend to experience.

JEL numbers: F41, E52

key words: central bank, developing, emerging markets, GDP, income, India, inflation, monetary policy, monsoon, nominal, shock, supply, target, terms of trade

The Best of Rules and Discretion: A Case for Nominal GDP Targeting in India

India’s central bank is contemplating a move from its current multi-indicator monetary policy approach, towards a simple credibility-enhancing nominal rule. As of 2014, it seems to favor a flexible inflation target, which it hopes will help lower inflation expectations that have been high and sticky since 20101.

In the paper we evaluate whether nominal GDP (NGDP) might be a good alternative for a monetary policy target for India, especially in the face of supply shocks that it faces and the varying macroeconomic situations that it experiences2. We outline a simple model to compare NGDP targeting with other nominal rules. We find that under certain simplifying assumptions and plausible conditions, NGDP targeting can indeed dominate Inflation targeting (IT).

Exhibit 1: Elevated inflation expectations in India

The paper is organized as follows. Section I discusses the origins and resurgence of NGDP targeting. The proposal has surfaced several times in the last few decades – though not always as a solution to the same probem -- suggesting its all-weather-friend characteristics. Section II highlights a simple theoretical model which compares alternative nominal monetary policy rules in terms of their ability to minimize a quadratic loss function that captures the objectives of price stability and output stability. We set out the conditions necessary for one regime to dominate the other. Section III discusses the evolution of monetary policy in India since independence in 1947 and the central bank’s desire to adopt a nominal rule. Section IV visits the different supply side shocks to which the country seems susceptible and the appropriateness of NGDP targeting to deal with them. Section V empirically tests the conditions outlined in section II to ascertain if indeed NGDP targeting would be appropriate in India in the sense of minimizing the quadratic loss function. We use two-stage-least-squares to estimate the parameters of the supply curve. Section VI addresses some practical concerns in implementing a NGDP rule. Operational complications arise from revisions in the nominal GDP statistics; but we find that inflation targeting has similar limitations. Section VIII concludes with thoughts on further research.

  1. Origins and resurgence of NGDP targeting and relevance for developing countries:

The earliest proponents of nominal GDP targeting were Meade (1978) and Tobin (1980), followed by other economists in the 1980s. The historical context was a desire to earn credible monetary discipline and lower inflation rates. The early 1980s saw monetarism as the official policy regime in some major countries. It was soon frustrated, however by an unstable money demand function.

NGDP targeting had been designed specifically to counter such velocity shocks. Nevertheless it was not adopted. The concept was on the backburner for several decades. Instead, the dominant approach for many smaller and developing countries between the mid-1980s and mid-1990s was a return to exchange rate targets.

A series of speculative attacks in the late 1990s forced many countries to abandon exchange rate anchors and move to some form of floating exchange rates. Another reason mid-sized open countries may want to have a floating exchange rate is to accommodate terms of trade shocks and other real shocks. But if the exchange rate is not to be the anchor for monetary policy, what is? The 2000s saw the spread of IT from some advanced economies to many emerging market countries. Over the last two decades, it is believed to have contributed to bringing down inflation across many countries and to have anchored expectations.

The Global Financial Crisis (GFC) of 2008-09 provoked concerns over shortcomings of IT, analogous perhaps to the frustrations with exchange rate targets that had resulted from the currency crises of the 1990s. Criticisms of Inflation Targeting include its narrow focus, lack of attention to asset market bubbles, failure to hit announced targets, and mistaken tightening in response to supply shocks such as the mid-2008 oil price spike.

Meanwhile, interest in NGDP targeting revived, as an alternative to inflation targeting3. But it has been focused on advanced economies such as the US, UK, Japan and Euroland, where interest rates have been constrained by ‘zero lower bound’. The motive for NGDP targeting in this literature is to achieve a credible monetary expansion and higher inflation rates, which are quite the opposite of what Meade (1978) and Tobin (1980) had in mind. This flexibility of NGDP targeting, as a practical way to achieve the goal of the day, be it monetary easing or tightening, and its focus on stabilizing demand are longstanding advantages.

Attention to developing and middle income countries in the NGDP targeting literature is scant4. And yet they may be the ones who can benefit the most from an NGDP target.

Developing countries have some characteristics that differ from advanced countries when it comes to setting monetary policy5. First, many developing countries have more acute need of monetary policy credibility. Some are newly born with an absence of well-established institutions, some have recently moved to a new monetary policy setting, some have a checkered past with central banks accommodating government debt, and some have had periods of hyperinflation.

The need for credibility amplifies the need to choose a nominal target which the central bank does not keep missing repeatedly. Central banks should choose targets ex ante that they will be willing to live with ex post and that they have relatively higher ability to achieve. For instance, announcing a strict inflation target that is then repeatedly missed would tend to erode credibility. One study showed that IT central banks in emerging market countries miss declared targets by much more than do industrialized countries (Fraga, Goldfajn and Minella, 2003).6

Second, developing and middle income countries tend to be more exposed to trade shocks (because they are more likely to export commodities and be price takers on world markets) and supply shocks (because of the importance of agriculture, social instability and productivity changes). Productivity shocks are likely to be larger in developing countries: during a boom, the country does not know in real time whether rapid growth is a permanent increase in productivity growth (it is the next Asian tiger) or temporary (the result of a transitory fluctuation in commodity markets or domestic demand).7

Weather disasters and terms of trade shocks are particularly useful from an econometric viewpoint, because these supply shocks are both exogenous and measureable. Exhibit 2 below shows that they tend to be bigger in emerging markets and low-income countries than in advanced economies.

The best choice of target variable depends on the type of shock to which the country is susceptible. If supply shocks are rife, NGDP targeting may be the right prescription. It splits the effects between inflation and GDP growth rather than suffering adverse supply shocks in the form of lost GDP alone.

Three categories of supply or trade shocks are relevant in particular:

  1. Pure supply shock – Natural disasters (such as an earthquake, hurricane, cyclone, tidal wave, or flood), other weather-related events (drought or severe winter), social disruption (labor strike or social unrest), and other productivity shocks (technological progress) fall under this category. For India, a poor monsoon is a good example. A fixed exchange rate by definition prevents the currency from depreciating and thereby moderating the fall in the trade balance and GDP. A CPI target, if interpreted literally, implies that monetary policy must be tightened enough to choke off any increase in the price level, leading again to lower GDP growth. Only in case of NGDP targeting can the currency respond to an adverse supply or terms of trade shocks by depreciating, helping the trade balance and splitting the adverse impact of the shock between inflation and growth.

  1. Rise in import price. One form of terms of trade shocks is an increase in the world price of importable goods. For India, oil prices are a good example. In the case of an exchange rate target, by definition the currency is prevented from depreciating, with adverse implications for the trade balance and GDP. In the case of CPI targeting, if interpreted literally, the currency must actually appreciate to prevent a rise in CPI inflation, with even worse implications for the trade balance and growth. In case of NGDP targeting, by contrast, the currency is not led to appreciate. Again the adverse shock is split between between inflation and GDP growth rather than growth alone.

  1. Fall in export price. Many developing countries export commodities that undergo large price swings on world markets. In the case of exchange rate targeting the currency cannot adjust. In the case of CPI targeting, depreciation is also limited as it would boost inflation. Thus the trade balance worsens, as does GDP growth. For the nominal GDP target, the currency depreciates, helping the trade balance as well as GDP growth to improve and moderating the economic contraction.

Exhibit 2: Emerging markets and low income countries are more susceptible to supply shocks

Source: IMF, 2011

Some theoretical models developed originally for advanced countries do not allow for comparing nominal rules in the face of exogenous balance of payments shocks. The advanced country models tend to assume that financing temporary trade deficits internationally is not a problem as international capital markets function well enough to smooth consumption in the face of external shocks. However, it is well documented by now that for developing countries, international capital markets can indeed exacerbate external shocks or even start them off. Capital inflow booms followed by sudden stops, sharp reversals, sharp depreciation and prolonged recessions are common in developing markets8.

NGDP targeting could potentially bring in the benefits of discretion without incurring the cost of inflation bias. In the model below, we formalize this intuition and probe further on when an NGDP target would dominate.

  1. Theoretical underpinnings: Comparing discretion with four alternative nominal rules:

It is well established that a credible nominal target eliminates the inflationary bias that discretion otherwise allows in a Barro-Gordon (1983) type model of dynamic inconsistency. But it makes a difference what is the choice of nominal target, because the economy is vulnerable to short run shocks (Rogoff, 1985; Fischer, 1990). The impact of the shocks depends on the variable chosen to be the nominal target.

Using a simple model outlined by Frankel (1995a,b)9, we compare five alternative nominal policy rules in the conduct of monetary policy: full discretion by the central banker, rigid money supply rule, rigid exchange rate rule, rigid price level rule and rigid nominal GDP rule, respectively.

Our investigation of these policy rules is predicated on the argument that one wants to announce some simple variable to which the central bank will commit. Credible commitment to a nominal target, for example, is a means of defeating the inflation bias from Barro-Gordon dynamic inconsistency.  The desire for transparency and accurate communication is not limited, however, to the Barro-Gordon argument for credible commitment to disinflation. It includes also the recent arguments for credible commitment to higher inflation. It is not even limited to a choice among alternative nominal anchors like M1 or inflation; the desire to offer forward guidance has included other intermediate targets such as the unemployment rate.10

The various Taylor rules are not members of this set of nominal-target commitments.   One could view the Taylor rules as writing down what discretion would do.  Or, as in "Flexible Inflation Targeting," one could view them as an intermediate (say monthly) means toward achieving a one-year or two-year target.  Either way, Taylor rules do not compete directly with IT, NGDP, etc.

Our model includes the money supply, an important variable in our view in thinking about the question of sterilization and allowing authorities the possibility to affect exchange rates. This is another departure from some of the advanced country models that have done away with the money supply on the grounds that money demand is unstable and central banks tend to use interest rates as their main policy instruments.

Inflation targeting and NGDP targeting can be formulated in terms of either levels or rates of change. A possible advantage of targeting a level is a faster return to the goal. If the regime is credible, an incipient shortfall in the NGDP level during the course of the year engenders expectations of a coming monetary expansion and higher inflation, thus contributing to lower real interest rates and an accelerated move towards the goal.11 The disadvantage of targeting a level is that the public may not fully understand, comprehend or believe a target in levels as it would a target in growth rates. For our model, the distinction between levels and growth may not be important. Targets are set each year; announcing it in levels or growth would amount to the same thing.

While we show derivations for different nominal rules here, our primary question of interest is CPI inflation targeting versus NGDP targeting. The former is the rule the RBI is currently considering and the latter is the rule we focus on.

To simplify the analysis, we assume rigid rules in our theoretical analysis, keeping in mind that welfare ranking for rigid rules in theory might be different than that for flexible rules (Rogoff, 1985). We start with the closed economy model and then move on to an open economy set-up.

The aggregate supply relationship is assumed to be:

  1. y = ȳ + b(p – pe) + u,

where y is real output, ȳ is potential output, p is the price index, pe is the expected price index (all variables can be in log levels or annual growth rate), and u is a supply disturbance.

  1. Closed economy objective function:

We assume objectives captured by the quadratic loss function:

  1. L = ap2 + (y – ŷ)2 ,

where a is the weight assigned to the inflation objective and ŷ is the desired level of output.

In order to build an expansionary bias to discretionary policy making, the ŷ > ȳ condition is imposed, as in Barro and Gordon (1983). For simplicity we have assumed that the preferred level of inflation is zero. Substituting (1) into (2),

  1. L = ap2 + [ȳ - ŷ + b(p – pe) + u]2 .

  1. Discretionary policy:

Under full discretion, the policy maker chooses aggregate demand so as to minimize the loss function every period (i.e. dL/dp = 0) giving:

  1. p = [- b(ȳ - ŷ) + b2pe – bu] / [a + b2].

Under rational expectations,

  1. pe = Ep =(ŷ - ȳ)b/a.

This term reflects the inflationary bias that Barro and Gordon (1983) attribute to discretion. Central banks have to inflate just to keep up with expectations, even without achieving higher output. The aim of a credible nominal target is to remove this inflationary bias. However, the economy will still be vulnerable to short run shocks, the impact of which will depend on the variable it chooses as the nominal target (Rogoff, 1985; Fischer, 1990).

Combining (5) and (4) gives the solution for inflation under discretion:

  1. p = (ŷ - ȳ) [b/a] – ub/[a+b2].

Combining (6) and (2) gives the value of the expected loss function:

  1. EL = (1 + b2/a)(ȳ - ŷ)2 + [a/(a + b2)] var(u)

The first term represents the inflationary bias while the second represents the impact of supply disturbance after authorities have chosen the optimal split between inflation and output.

  1. Money rule:

The money market equilibrium condition is given as –

  1. m = p + y – v,

where v represents velocity shocks. We assume that v is uncorrelated with u. If authorities pre-commit to a money growth rule to reduce expected inflation in the long run equilibrium, they must give up on affecting y. The optimal money growth rate is the one that sets Ep = 0; thus setting money supply, m, at Ey, which in this case is ȳ. The aggregate demand equation thus becomes:

  1. p + y = ȳ + v.

Combining (9) with (1) gives,

  1. y = ȳ + (u + bv)/(1+b), p = (v-u)/(1+b).

Substituting into (2) gives,

  1. EL = (ȳ - ŷ)2 + {(1+a)var(u) + (a+b2)var(v)}/(1+b)2

The first term is smaller than the corresponding term in the discretion case, because pre-commitment eliminates expected inflation. But the second term could likely be larger, as the authorities give up the ability to respond to money demand shocks. Which regime is better, depends on the size of the shocks and the value of a, i.e., weight placed on price stability.

  1. Nominal GDP rule

In the case of a nominal GDP rule, authorities vary money supply in such a way that velocity shocks are accommodated and p+y in equation (9) is constant. The solution is the same as in the money rule but with the v disturbance dropped out. As such, the loss function is reduced to

  1. EL = (ȳ - ŷ)2 + [(1+a)/(1+b)2] var(u).

This unambiguously dominates the money rule (11).

It is not possible to know whether the rule dominates discretion unless the values of the parameters such as var(u) and a are known. While the first term (reflecting inflationary bias) is smaller for the nominal GDP rule, the second term depends on the value of the parameters (which we estimate later in the paper).

  1. Inflation rule:

The authorities set monetary policy so that the price index (level or rate) is zero, not just expectation but also regardless of later shocks. From (3) this gives

  1. L = [(ȳ - ŷ) + u]2

  1. EL = (ȳ - ŷ)2 + var(u).

Comparisons show that the price level rule dominates the money supply rule if velocity shocks are large. If they are small, the money supply rule collapses to the nominal GDP rule.

The nominal GDP rule dominates the price rule if –

[(1+a) / (1+b)2] < 1, i.e., so long as a/b < 2 + b.

The conclusion is NGDP targeting dominates IT except in the presence of a very steep supply curve and/or a very high weight on the price stability objective in the quadratic loss function.

The condition can be simplified further if one is willing to infer an estimate a, the weight on price stability, from the Taylor Rule. The original Taylor Rule, which is still widely used, gives equal weights to output and price stability in setting its real interest rate policy instrument, implying a = 1. Then the condition a/b < 2 + b collapses to b > – 1. This implies that for the nominal GDP rule to dominate the price rule, the AS curve must be flat enough that its slope (1/b) is less than 2.414.

  1. Open economy objective function:

The conduct of monetary policy in an open economy accounts for the exchange rate. To account for the central bank’s objective of having a stable exchange rate, we include it directly in the loss function along with output and inflation.12 The intention is to err on the side of exchange rate targeting and to understate the case for NGP targeting.

The open economy loss function (equivalent to equation 2 from above) now looks as follows:

  1. L = ap2 + (y – ŷ)2 + cs2 ,

Where s is the spot rate measured relative to some equilibrium or target value and c is the weight placed on exchange rate stability. We assume here that the central bank would like to minimize volatility in the exchange rate around its average value.

Rather than specifying a full-fledged model for a free floating exchange rate, we attribute the bulk of the variation to an error term that we call e here. This is in line with the fact that much of the variation in exchange rate (even ex post) is hard to explain. We also include the money supply, to allow authorities the possibility of affecting the exchange rate. (If the exchange rate is fixed, then e can be interpreted as a balance of payments shock, which shows up in the money supply.) Our equation is as follows:

  1   2   3

The database is protected by copyright © 2016
send message

    Main page