Discussion

Simon van Norden

In October 1993, the Bank of Canada held a research conference onthe economic implications of low inflation. In his introduction to theresulting conference volume, Stephen Poloz (then head of the Bank’sResearch Department) noted that at least two topics had been consciouslyomitted from the conference agenda: (1) measurement bias in the consumerprice index, on which an extensive technical report had just been published(Crawford 1993); and (2) the empirical link between inflation and economicperformance. On the second point, Poloz (1994, ii) wrote:

There already exists a great deal of published work on thisissue, and more recent contributions to this literature haveshown that previous estimates of this link based on cross-country evidence are very fragile.

The conference volume was subsequently reviewed in theCanadianJournal of Economics, where Johnson (1995, 724) describes this secondomission as

what I feel is the biggest disappointment of the volume….[T]his topic is much too important to be omitted; in thevolume… Howitt (1990) [is quoted] as saying this topic “maybe the most important issue of all.” I agree.

Noting that other papers in the volume conclude that the welfarebenefits of low inflation are potentially large because theyassume that lowerinflation raises labour productivity growth, Johnson (ibid.) goes on to say:

117

118Discussion: van Norden

The link between inflation and productivity growth is tooimportant not to be addressed directly in the volume; if thelink is so fragile, its fragility should have been reportedclearly, so that readers would have the information to assess itsfragility.

He then concludes (ibid., 725-26):

If the current regime is permanent, we shall experience a longperiod of low inflation and discover if the benefits of lowinflation are significant. Another conference could theninvestigate the issue and produce another set of papers.

How could anyone resist such an invitation? It should therefore comeas no surprise that in this 1997 conference the Bank has consciously decidedto examine this question, both in the Ambler and Cardia paper, whichaddresses the empirical literature, and in the Black, Coletti, and Monnierpaper, which asks how large the long-run benefits of low inflation must be tojustify the short-run costs of lowering inflation. Black, Coletti, and Monnieralso survey some of the literature.

The intervening years have shown that this topic is far fromexhausted, to judge by the number of publications on it. One year alone sawinteresting papers on the subject by Sarel (1996), Barro (1996), Judson andOrphanides (1996), Hess and Morris (1996), Dotsey and Ireland (1996), andCameron, Hum, and Simpson (1996), among others. If anything, publicationrates on this topic may not yet have crested.

In approaching this still rapidly evolving literature, Ambler andCardia chose not to give us a synthesis or a snapshot of the empiricalevidence. Their paper is more ambitious; it tries to shape our thinking aboutthe relationship between this empirical literature and the policy question athand. What I think of as the most central of their many contributions can besummarized as follows:•

Estimates of the correlation between inflation and output growth willbe a combination of a non-monetary effect and the effect (if any) dueto a change in monetary policy. The relative weights of these twoeffects depend on the relative importance of monetary and non-monetary shocks in our data. However, only the effect due tomonetary policy is relevant when central bankers are consideringwhether to pursue low-inflation policies. Put algebraically, when we

˙, on inflation,π, we getregress output growth,y

ˆ+ε, andˆ+π•βy˙=α

ˆ=γ•ββmonetary+(1–γ)•βnon-monetary.

(1)(2)

Discussion: van Norden119

The challenge for those seeking to guide policy is to disentanglethese two effects.•

In a world with exogenous money growth and stable velocity, non-monetary shocks create a strong negative correlation betweeninflation and real growth, soβnon-monetary=–1. To see this, recallthe identity

MV≡PY, som˙+v˙≡π+y˙,

(3)

˙ andv˙ are, respectively, the rates of change of the moneywherem

stock, and velocity.

d(m˙+v˙)dy˙

-=–1.If--------------------=0, then-----dy˙dπ

(4)

In effect, this implies that policy is equivalent to a nominal income-targeting rule; since the growth rate of nominal income is fixed, it

follows that higher real growth is associated with a one-for-onedecrease in inflation.••

Ambler and Cardia’s model generates a very slightly negative value

˙⁄dπ when only monetary shocks are present.fordy

For the standard mix of shocks, it generates a negative relationshipthat is much stronger than that found in the data (−0.22 in the dataversus−0.8 in the model).

To understand the relevance of their model for policy, I think we needto understand why it falls so far short of capturing what we see in the data.Put another way, the contentious issue may not be whether the correlation inthe data is different from zero, but why it is not closer to−1. There are fourpossibilities to consider:

ˆ is mismeasured;1.β2.γ is much larger than in their model (that is, monetary shocks aremuch more important);3.βmonetary is a big positive number; and4.βnon-monetary is closer to zero.

With respect to point (1), Ambler and Cardia suggest that the Phillipscurve could be an important source of the mismeasurement. Specifically, ashort-run positive relationship between inflation and growth would tend toraise their estimated covariance of growth and inflation, since their estimateis based on unfiltered data.

Ambler and Cardia explicitly consider the possibility of point (2) ascase 4 in their Table2, where they set new parameters for the model withmonetary shocks that are 10 times more important than in their base case.

120Discussion: van Norden

Even under these extreme assumptions, the simulatedβ is, in absolutemagnitude, roughly double what we get from the data. It therefore seemsthat changes inγ alone cannot explain the difference.

To digress somewhat, a potentially interesting empirical puzzlearises from their assertion that increasing the importance of monetaryshocks should raise (that is, make closer to zero) our estimate ofβ. Somerecent work on the inflation-to-growth relationship (Sarel 1996, Bullardand Keating 1995, Barro 1996, Judson and Orphanides 1996) argues that

ˆ is more negative) among high-inflationthe relationship is stronger (thatβ

countries than among low-inflation countries. However, it is presumably inthese high-inflation countries where the variance of monetary shocks ismore important and where we should therefore expect a weakerrelationship. I would be interested to see a more thorough consideration ofthis puzzle.

Another explanation, point (3) above, would be that their modelgets the sign wrong for the trade-off between monetary inflation andgrowth. Presumably, if higher long-run inflation leads to higher (notlower) long-run growth, then the mixture of the two effects we see in thedata will be closer to zero than their model predicts. They mention oneexample of this—that is, a model where investment in human capital is asubstitute rather than a complement to labour-market activity; another isthe case where the loss in government seigniorage revenue from lowerinflation must be recouped through alternative taxes that are moredistortionary than the inflation tax. However, the problem is that theeffects of inflation on labour-market activity in their model are so weak,and seigniorage revenue so small, that it is hard to believe that either islikely to get us close to the values we see in the actual data. What we seemto need for this kind of explanation is a mechanism whereby a low-inflation policy has much larger permanent negative effects on growth. Inthis light, it is interesting to consider the evidence for downward nominalwage rigidity as discussed in this conference’s third session, and whetherit could give a large enough effect.

Finally, point (4) suggests thatβnon-monetary might be much closer tozero if the authors’ exogeneity assumptions are unrealistic, which I suspectis the case. To derive the coefficient of−1 for non-monetary shocks, we hadto assume that such shocks affected neither money growth nor velocity.Suppose we instead assume that monetary policy will react to exogenousnon-monetary shocks, although the reaction might change across time andnations. The resulting relationship is just

˙dy˙dm˙dv

-----=------+-----–1.dπdπdπ

(5)

Discussion: van Norden121

Ambler and Cardia argue thatdv˙⁄dπ is close to zero.1 This meansthat, to explain the results in the data, we need a monetary policy thattends to accentuate changes in inflation due to non-monetary shocks(dm˙⁄dπ>0). This may be the case if, for example, monetary authorities tryto run countercyclical policies but take time to learn about changes in thelevel of potential output. Many observers think this is a goodcharacterization of monetary policy responses to oil price shocks in the1970s.

I hope that these considerations of the possible effects of downwardnominal wage rigidity, Phillips curves, and monetary policy reactions to realshocks will be persuasive that there are several interesting and credibleexplanations for the gap between the authors’ simulated values and theestimates ofβ, and that these explanations deserve additional carefulscrutiny and comparison.

In these remarks, I have assumed that the main point of the Amblerand Cardia paper was to emphasize the distinction between the growth-inflation correlation and the trade-off between the two that is exploitable bypolicymakers. Others have interpreted the paper quite differently andconcluded that the authors’ main message was that the exploitable trade-offfor policy is very, very small. While this is certainly the result that theirmodel produces, I wish to explain why I do not see that as a centralconclusion.

I think that Ambler and Cardia are careful to present their model asbeing simply one of many possible ways of modelling the effects of inflationin an endogenous growth framework. They make no claims that their modelshould give an upper or lower bound to the size of the policy trade-off, orthat all reasonable models will give similar results. It seems to be generallyaccepted, for example, that the interactions of inflation with an imperfectlyindexed tax system can increase the after-tax cost of investment, and that forreasonable parameters such effects are much larger than those produced bythe “consumption tax” mechanism Ambler and Cardia consider. True, tax

1.The thought experiment they consider is a one-time permanent shift in the level ofinflation. We should not expect this to permanently affect thetrend growth rate of velocity,although it may well permanently shift thelevel of velocity. However, it seems to me thatthis is not quite consistent with the authors’ discussion of the time-series literature oninflation and growth. There they argue that historical inflation should be treated as astationary variable. While I am interested in the econometric value of their arguments, Ileave that for others to discuss. Instead, let me ask what is the possible relevance ofquestions about the effects of long-run changes in inflation if inflation is stationary, so thatthere can be no long-run changes? What precisely is the mechanism that prevents monetaryauthorities from influencing inflation rates in the long run? Does this imply that our currentemphasis on low inflation can be nothing but a transitory policy fad?

122Discussion: van Norden

effects are problematic for those seeking to justify a move to low inflation:Why not just reform the tax system? What about the costs of forgone taxrevenue? Such questions deserve careful consideration. However, my pointis simply that it is wrong to cite this paper as “proof” that the growth effectsof a low-inflation policy cannot be large.

To summarize, Ambler and Cardia make an interesting contributionto the empirical literature on the relationship between inflation and growth. Ithink their most valuable message is that we must not confuse thecorrelation (conditional or unconditional) between these series with therelevant trade-off facing policymakers. While this paper does not yet allowus to answer Johnson’s question—whether there is a link between inflationand economic performance—it makes me hopeful that subsequent work willgive us a more complete understanding.

References

Barro, R.J. 1996. “Inflation and Growth.”Federal Reserve Bank of St. Louis Review 78 (May–June):

153-69.

Black, R., D. Coletti, and S. Monnier. 1997. “On the Costs and Benefits of Price Stability.” Bank of

Canada. This volume.

Bullard, J., and J.W. Keating. 1995. “The Long-Run Relationship between Inflation and Output in

Postwar Economies.”Journal of Monetary Economics36 (December): 477-96.

Cameron, N., D. Hum, and W. Simpson. 1996. “Stylized Facts and Stylized Illusions: Inflation and

Productivity Revisited.”Canadian Journal of Economics 24 (February): 152-62.

Crawford, A. 1993.Measurement Biases in the Canadian CPI. Technical Report No. . Ottawa:

Bank of Canada.

Dotsey, M., and P. Ireland. 1996. “The Welfare Cost of Inflation in General Equilibrium.”Journal of

Monetary Economics 37 (February): 29-47.

Hess, G.D., and C.S. Morris. 1996. “The Long-Run Costs of Moderate Inflation.”Federal Reserve

Bank of Kansas City Economic Review81 (1996Q2): 71-88.

Howitt, P. 1990. “Zero Inflation as a Long-Term Target for Monetary Policy.” InZero Inflation: The

Goal of Price Stability, edited by R.G. Lipsey, 67-108. Toronto: C.D. Howe Institute.

Johnson, D. 1995. “Economic Behaviour and Policy Choice under Price Stability: Proceedings of a

Conference Held at the Bank of Canada, October 1993.”Canadian Journal of Economics28 (August): 723-26.

Judson, R., and A. Orphanides. 1996. “Inflation, Volatility and Growth.” Finance and Economics

Discussion Series No. 96-19. U.S. Board of Governors of the Federal Reserve System,Washington, D.C.

Poloz, S.S. 1994. “Introduction.” InEconomic Behaviour and Policy Choice Under Price Stability,

i-iv.Proceedings of a conference held at the Bank of Canada, October 1993. Ottawa: Bankof Canada.

Sarel, M. 1996. “Nonlinear Effects of Inflation on Economic Growth.”International Monetary Fund

Staff Papers 43 (March): 199-215.