A. W. Phillips - Employment, Inflation And Growth

The Suntory and Toyota International Centres for Economics and Related Disciplines

Employment, Inflation and Growth Author(s): A. W. Phillips Source: Economica, New Series, Vol. 29, No. 113 (Feb., 1962), pp. 1-16 Published by: Blackwell Publishing on behalf of The London School of Economics and Political Science and The Suntory and Toyota International Centres for Economics and Related Disciplines Stable URL: http://www.jstor.org/stable/2601516 Accessed: 08/11/2008 10:51 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=black. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

The London School of Economics and Political Science, Blackwell Publishing, The Suntory and Toyota International Centres for Economics and Related Disciplines are collaborating with JSTOR to digitize, preserve and extend access to Economica.

http://www.jstor.org

1962]

Employment, Inflation and Growth' By A. W. PHILLIPS 1.

INTRODUCTION

Since the end of the second world war economic policy and controversyin Britainhave been directedto ways of attaininga number of related objectives,prominentamong which are the maintenance of a high and stable level of employment,reasonablestabilityof the averagelevel of final productprices, a fairly rapid rate of economic growth,a satisfactorybalanceof foreigntradeand reasonablestability of foreign exchangerates. It can hardly be claimed that there has been complete success in the attainmentof these objectives. It is true that employmenthas been maintainedat an extremelyhigh level. The averagelevel of unemploymentsince 1948 has been little more than 12 per cent. of the work force. This is probablya lower level of unemploymentthan that attainedin any single year in peace time during the previous century, except perhapsin 1872. Even in the boom yearsof the pre-wartradecyclethe percentageof tradeunionists unemployedrarelyfell below 2 per cent. and over the 53 years from 1861 to 1913it averaged42 per cent. The actions taken to improve the stability of the system have also had some measureof success. Beforethe first world war there was a fairly regulartrade cycle with an averageperiod of about eight years, during which trade union unemploymentfluctuatedbetween about 2 and 10 per cent. Since the second world war the cyclical movementsin economic activity have becomemore rapid,with a period of four or five years, but the fluctuationshave been smaller. Unemployment,indeed, has only fluctuatedbetweenabout 1 per cent. and 21 per cent., that is, over a range of about 12 per cent., but the percentagefluctuationsin gross nationalproductabout the growth trend have been about five times as large as this,2the range of the fluctuationsas a percentageof the trend being about 7 or 8 per cent. Between 1948 and 1960, gross national productincreasedat an averagerate of about 2i per cent. per annum and productivityper man hour at perhaps 1l per cent. per annum. Thoughthese rates of increaseprobablycomparefavourably with those in earlierperiods of British history they are lower than those of a number of other industrialcountries in the same period. The averagerate of rise of the retailpriceindex between1948 and 1960was 3.7 per cent. per annum. Therewould be fairlygeneral agreementthat this rate of inflationis undesirable.It has undoubtedly 1 Inaugural lecture given at the London School of Economics and Political Science on 28th November, 1961. 2 See F. W. Paish, " Output, Inflation and Growth ", to be published. 1

A

2

ECONOMICA

[FEBRUARY

beena majorcauseof the generalweaknessof the balanceof payments and the foreign reserves,and if continuedit would almost certainly make the presentrate of exchangeuntenable. It is my belief that one of the main reasonsfor the difficultiesthat have been experiencedin devising and implementingappropriate economic policies is lack of adequate quantitativeknowledge and understandingof how the economicsystemworks. Of course, economists do understand,in a generalsort of way, quite a lot about the workingof the economicsystemand do now have a mass of quantitative informationabout importanteconomic variables. But in order to bring this knowledgeto bear on the problem of formulatingand attaininga consistentset of policyobjectiveswe requirealso knowledge of the quantitativerelationsbetweeneconomicvariables. In particular it is necessaryto know what quantitativerelationshold betweenthose economic variableswhich are either the objectivesof policy or the instrumentsthroughwhich we attempt to attain the objectives. For example,if some relationholds, in given institutionalconditionsand on averageover a period of years, betweenthe level of employment and the speed of inflation,failureto take account of it may lead to the adoptionof inconsistentobjectivesand to a type of schizophrenic behaviouras attempts are made to attain these inconsistentaims. Knowledgeof the relation would lead either to modificationof the objectivesto make them consistent or perhaps, since an economic relationis only the resultof fairlyregularpatternsof humanbehaviour, to some modificationof institutionsor behaviourwhich would alter the relation so as to permit some more desirablecombinationsof consistentaims. Or again, if at a certaintime unemploymentis felt to be too high and short-terminterestrates are loweredin order to raise the demand for goods and so for labour, how large will the effects be and when will they occur; will the higher demand also lead to an increasein fixedinvestmentand if so how large an increase and afterwhat intervalof time; will wage rates and pricesrise more rapidly as a result of the higher demand; if internal demand and pricesrise will importsrise and exportsfall, and if so when and by how much? If we are to assessthe effectsof our attemptsto influence the course of economic affairswe need answers,numericalanswers, to questionslike these. If we do not have this knowledgethe policy adjustmentswill almost certainlybe inappropriatein magnitudeor timingor both and may well cause,as I believethey have often caused in the past, unnecessaryand harmfulfluctuationsin economicactivity. We may obtain tentative estimates of a quantitativerelation in economics by making a preliminarysubjectiveanalysis of human motivationand behaviourand then carryingout a statisticalanalysis of relevantdata from past records. We also need to investigatethe degreeof errorthat theremay be in the estimate,the extentto which the relationin successiveshort time periodsdepartsfrom its average overlongerperiodsand whetherthereis any evidencethat the average

1962]

EMPLOYMENT,INFLATIONAND GROWTH

3

relation changesin any systematicway throughtime. On the basis of quantitativeestimatesof this sort it is possibleto set up simplified models of the economic system and by studying the propertiesof differentmodels with a variety of policy relationshipswe can form some judgementof the likely effectsof alternativetypes of economic policy. The empiricalstudyof economicrelationsand the quantitative investigationof the behaviourof models of economic processesare comparativelyrecent developmentsin our subject. The knowledge and understandingwhich have so far been gained are far from being adequate for a firm and detailed appraisalof economic policy. I think they do, however,make possible some attemptat clarification of the problemswe face and justify some suggestionsfor methodsof dealingwith them. 2.

SOME GENERAL PRINCIPLES OF FLUCTUATIONS AND STABILITY

The first policy objectiveI should like to consideris that of maintaining a stable level of employment; but before proceedingto this it seems desirableto illustratesome generalpoints about equilibrium, stabilityand fluctuationsin the simplestpossibleway. Let me therefore consider a single commoditywhich is being continuouslyproduced and consumedand whichis tradedon a perfectlycompetitivemarket. Assumethat the rate of productionis an increasingfunctionof price, the rate of consumptionis a decreasingfunctionof price and that the rate of changeof the price is proportionalto the excess demand,i.e., to the rate of consumptionminusthe rate of production. This is the simpletext-bookexampleof supplyand demandin a singlecompetitive market. It is frequentlystated,and has indeedbeen statedwith some emphasisby sucheminenteconomistsas Walras,MarshallandWicksell, that such a systemis necessarilystable,i.e., that it alwaystends to an equilibriumin whichthe priceis suchthat the ratesof productionand consumptionare equal. The argumentis usuallyvery simple. Suppose the systemis not in equilibrium;for example,supposethereis excess demand. Then the price will rise. This will increaseproductionand reduceconsumptionand so reducethe excessdemand. Sincethe price continuesto rise so long as there is any excess demandand any rise in price reduces the excess demand the process will continue until the excess demandis eliminated. In brief; the existenceof any discrepancybetween productionand consumptioncauses a movement in price which tends to correctthe discrepancy. Therefore,the argumentruns,the systemis stable. This argumentis, of course,fallacious except on the assumptionthat the completeresponseof the rates of productionand consumptionto any change in price occurs instantaneously. If thereare any time lags in any of the responsesthe system willusuallyfluctuate.Whetherthe fluctuationswill die awayor whether they will increasein amplitudeand tend to some regularand sustained limit cycle dependson the preciseformsof the time lags, on the slopes

4

ECONOMICA

[FEBRUARY

of the supplyand demandcurvesand on the speed at whichthe price changeswhen there is a given excess demand. The competitivemodel which we have been considering,and other so-called" self-equilibrating"models of economic processesare, in fact, examplesof what are knownin otherfieldsof studyas " negative feed-backcontrol systems". In order to see intuitivelywhy these systemsare often oscillatoryand may well be unstable(whichmeans in practicethat they tend to producefairly large and regularcyclical movements)let us consider again the competitivemarket. Assume this timethat somefactorotherthanpricecausessmallcyclicalchanges in productionor consumption,so that excess demand is alternately positiveand negative(see curve a of Figure 1). We shall call this an exogenousmovementof excess demandand see what furthermovements in excess demandwould be induced by price changes which dependedonly on this exogenousmovement. In otherwordswe shall find what would usuallybe called the correctivemovementsin excess demandwhich result from the price changescaused by the exogenous movementsin excess demand. Since we are assumingthat the rate of changeof priceis proportionalto the exogenousmovementsin excess demand,the rate of changeof pricewould have the sametime pattern as the exogenousmovement. The price itself, however, would lag behindthe rate of changeof priceby a quarterof a cycle (see curveb)

a. Exogenous movement,and rate of change of price +

b. Price +

C.Induced movement,with no time lag +

d. Induced movement, with time lag Figure 1

1962]

EMPLOYMENT,INFLATIONAND GROWTH

5

since the price would be at its maximumwhen its rate of changewas zero and changingfrom positiveto negative,and would be at a minimum when its rate of changewas zero and changingfrom negativeto positive. If the completeresponseof productionand consumptionto any change in price was instantaneous,the excess demand induced by the movementsof price would be at its minimumwhen price was at its maximumand vice versa(see curvec). Suppose,however,that productionand consumptionrespondedto changes in price with a time lag equal to one quarterof the period of the cycle. Then the excessdemandinducedby the pricechangeswould be exactlyin phase with the exogenousmovementsin excessdemand(see curved). Instead of tendingto offset or correctthe exogenousmovement,the induced changeswould tend to accentuateor amplifythe fluctuationsin excess demandcausedby the exogenousmovement. It is intuitivelyplausible, and can in fact be proved,that if, in this case, what are usuallycalled the equilibratingor correctiveforces are strong enough to make the amplitudeof the inducedmovementsin excess demandgreaterthan the amplitudeof the exogenousmovements,the systemwill be unstable, that is, the fluctuationswill increaseand tend towardsa regularand sustainedlimit cycle. The exogenousmovementsor disturbanceswhich affect economic activityare not usually,of course,of the simpletype assumedin this example. They are more likely to be of a ratherarbitraryor random patternwhichcanbe describedonly in termsof a statisticalor stochastic process. Suchprocessescan,however,be analysed,by a methodknown as spectral density analysis, into cyclical componentswith periods rangingover the whole spectrumfrom zero to infinity. If disturbances of this sort operateon the modelwe havebeenconsidering,any cyclical componentwhose periodis such that the time lag in the responseof excess demand to price results in a lag in the neighbourhoodof a quarterof the cyclical period will be amplified,while cyclical componentswhose periodsare widely differentfrom this will be reduced in amplitude. If the " correctingforces" are sufficientlystrong the cyclicalcomponentswithina particularrangeof frequenciesor periods will be amplifiedto such an extentthat they will dominatethe market movements,which will then exhibit large and somewhat irregular fluctuations,in which, however, cycles with this particularrange of periodicitieswill predominate. Thereare two more mattersof some importancethat I should like to referto whiledealingwith the generalprinciplesof fluctuationsand stability. The first concernswhat I shall call the form or distribution of the time lags. Supposethe pricein the commoditymarketwe have been consideringwas constantfor a long time and then suddenlyrose and remainedconstantat the new level. The rate of productionwould eventuallyincreaseto some higher value, but there are any number of time pathsthat this increasemightfollow. For example,production might start to rise immediatelyand continue to rise at a gradually

6

ECONOMICA

[FEBRUARY

diminishingrate, or it might remain constant for a while and then increasegradually,or it might remainconstantfor a longer time and then increasevery suddenly(see curves a and b and c in Figure 2). Now it can be shown,but by methodswhichI shall not inflicton you here, that the existenceof a time lag of the firstform in the corrective process is much less likely to cause fluctuationsand instabilitythan a lag of the second form, and a lag of the second form is less likely

Price

c Production Figure

2

to cause these troublesthan is the third form of lag. An important rule in devisinga correctivesystem is thereforeto get the corrective adjustmentstartedimmediatelythe discrepancyit is intendedto correct beginsto be observed. Provideda fair proportionof the effectof the correctiveaction is obtainedfairly quicklyit does not much matter if the remainingeffectsare delayed; but it does mattervery much if the correctiveaction itself or all of its effectsare delayed. The secondmatterwhichI shouldlike to deal with brieflyis that of alternativetypes of correctiveaction. It is perhapsbest introducedby askingwhat would happen if in the simple commoditymodel which we consideredearlierit was the priceratherthan the rate of changeof pricewhich dependedon the excessdemand. We see from the curves that the price movementswould occur a quarterof a cycle earlieras a result of this modification,and the induced movementsin excess demandwould no longer be in phasewith the exogenousmovements, so that the destabilisingeffects of the inducedmovementswould be reducedor eliminated. Dependingon the form of the time lag, it might or might not be possible to find a cycle of higher frequency, that is, of shorterperiod,for which the lag in the responseof excess demandto pricewas half a cycle insteadof a quartercycle. If so, an exogenousmovementof this higherfrequencywouldlead to an induced movementin phasewiththe exogenousmovementand if the amplitude

19621

EMPLOYMENT,INFLATIONAND GROWTH

7

of the inducedmovementwas sufficientlylarge instabilitycould still occur at the higherfrequency. But since most forms of time lag are in effect weightedaveragesand tend to producea low amplitudeof responseto high frequencycycles the likelihoodof cyclicalinstability would be reduced. On the other hand this form of correction,which is called proportionalcontrol,could neverensurethat productionand consumptionwere brought into equality, since the price would be constant if excess demandwere constant, even if it were not zero. However, a combinationof proportionalcontrol and the type of correctionused before, which is called integral control, overcomes this difficultyand givesbetterperformancethan integralcontrolalone. We may go furtherin this directionand considerthe effectof adding to the correctiveaction a componentbased on the rate of change of excessdemand. Sincethe fluctuationsin the rateof changeof a cyclical variablelead the fluctuationsin the variableitself by a quarterof a cycle,this componentof correctiveaction,knownas derivativecontrol, has a somewhatsimilareffect to that which would be obtained by basingthe correctiveaction on a forecastof excessdemand. A derivative componentof controlis used in combinationwith proportional and integralcomponentsin most negativefeed-backcontrol systems. By an appropriatecombinationof the three componentsit is usually possibleto obtainvery good regulatingperformanceof a system,with correctiveactions based only on the actual values of the variables and their rates of change in the immediatepast, and without any recourseto predictedvalues or forecasts. 3.

FLUCTUATIONS IN EMPLOYMENT AND ECONOMIC ACTIVITY

I hope I have not boredyou too muchby this ratherlong and somewhat technicaldigressionon the generalprinciplesof fluctuationsand stability. But I think some understandingof these principlesis helpful in a discussionof the stabilityof employmentand economicactivity. For the correctiveadjustmentswhich affect employmentand activity in the whole economy, whetherthey be inherentin the working of the economyor appliedas deliberateinstrumentsof policy, are again examplesof control by negativefeed-back. The use of forecastsin policy does not substantiallyaffectthis statement,since the forecasts are themselveslargelybased on observationsof the economy and its movementsin the recent past. Indeed,given the presentstate of the art of forecastingI believe that better results might be obtained by basing suitablecorrectiveaction directlyon observationsof the economy and its changesrather than on forecastswhich are themselves largelyderived,perhapsby dubiousprocesses,fromthoseobservations. Even if there were no specialfeatureswhich might accentuatedisturbancesand fluctuationsin the economyas a whole, it could not be assumed that the existence of corrective adjustments,even quite powerfulcorrectiveadjustments,would stabilisean economy. If they operate with long time lags, and especiallyif there are long delays

8

ECONOMICA

[FEBRUARY

beforethey commenceto operate,they will cause cyclicalfluctuations, and the strongerthe correctiveforces the more violent will be the fluctuations. But thereis fairlygeneralagreementamong economists that thereare specialfeaturesof an economicsystem,in particularthe multiplierprocess and adjustmentsin inventoriesand capital equipment,whichaccentuatedisturbancesand tend to causecyclicalfluctuations. These featuresincrease the need for deliberatestabilisation policies, but also make it more difficultto devise suitablepolicies. The problem is best studied by investigatingthe propertiesof a variety of models in which use is made of the limited amount of quantitativeknowledgeat presentavailableabout economicrelationships. The work involved in such investigationsis greatly reduced and the range of possible models is widenedby the use of modern electronicequipment. The results obtained in this way cannot, of course,be conclusive,but from the investigationsI have done so far I have considerableconfidencein two simple propositions. The first is that correctiveaction taken in an attemptto reducethe anmplitude of the short cycle of four or five yearswhichis typicalof the post-war period is not likely to be successfulunless it is based on recently observedrates of changeof economicactivityas well as on the level of activity. The otheris that evenif the correctiveactionis appropriate in this respectit will still be unsuccessfulunless a fair proportionof its ultimatedirecteffecton demand,say about a quarterof it, occurs withinthreeor fourmonthsof the occurrenceof the errorit is designed to correct, and at least one half of the full effect within about six months. Let us examinesome of the existingmeans of influencingdemand in the light of these requirements. The response of investmentto changesin monetaryconditionsand interestrates is almost certainly delayedand slow. There is probablya delay of some monthsbefore decisionsto invest are significantlyaffectedand with most types of investmentthere is probably a further long time lag between the decisions to invest and the actual productionof capital goods. If this is so, fluctuationsare likely to be intensifiedratherthan reduced by attemptsto correctthem throughoperatingon long-terminterest ratesand investmentin fixedcapital. This does not mean that interest rate policy is unimportant.I believeit has a vital role to play in the slower adjustmentsrequiredas a result of changes in the desire to save or invest,and thus in influencingthe averagelevel of employment andthe averagerateof changeof the priceleveloverfairlylong periods. It is more difficultto judge the effect of operatingon short-term credit and short interestrates. Adjustmentsin these can be made more quicklythan in long rates, and to the extentthat they affectthe desire or ability to hold inventoriesthey might have a significant effect on productionwithin two or three months. But I think much more empiricalwork will be needed before one can judge with confidencethe magnitude,speed and reliabilityof these effects. I suspect

1962]

EMPLOYMENT,INFLATIONAND GROWTH

9

that if fairly large and rapid adjustmentsin short-termcredit conditionsweremadein responseto both the level and the rate of change of economicactivitythey would help to reducethe amplitudeof the short cycle. But there is a difficultyin using adjustmentsof shorttermcreditandinterestratesfor this purpose. Shortrates,and perhaps to a lesser extentcredit,are closelyrelatedto bank rate, and the level of bank rate is often made to depend as much on the state of the foreignreservesas on the internalconditionof the economy. To the extentthat fluctuationsin the foreignreservesare the resultof fluctuations in the balanceof tradethey will tend to lag behindfluctuations in the balanceof trade by a quarterof a cycle. Since the balanceof trade moves fairly closely with internal activity, fluctuationsin the reservestend to lag behindinternalactivity,so that changesin bank rate tend to be too late for satisfactorycorrectionof economicfluctuations. My conclusionconcerningmonetarypolicy is thus similarto that of the Radcliffe Committee;' " . . . monetary measures cannot alone

be relied upon to keep in nice balancean economy subjectto major strainsfrom both without and within. Monetarymeasurescan help, but that is all ". My reasonsfor this conclusionare perhapsa little differentfromthose of the Committee.I thinkthat changesin interest rates and credit conditionsprobablydo have quite powerfuleffects on demand, but that their usefulness for correcting short-period fluctuationsis seriouslylimitedas a result of the long time lags in the responseof investmentto changesin interestrates, and in the case of credit as a result of the temptationor need to make the adjustments in responseto the state of the foreignreservesratherthan at the times appropriateto the correction of fluctuationsin internal economic activity. Much more researchwill have to be done, however,before the last word is said on these matters,and I would heartilyendorse the Committee's statement2 that " . . . it is essential to have much

greaterand more systematicknowledgeof the factors that make up the financialsystemand of theirrelativemovements". Turningto fiscal methods of influencingdemand, it is clear that adjustmentsthroughannual budgetsdo not meet the requirementsI have stated for the correctionof short-periodcycles. Budgetchanges have probablyplayed some part, together with monetarypolicy, in overcomingthe more severeand longercyclesof pre-wardays, though this may have come about as muchthroughthe confidenceof business men that the budgetcould and would be used to averta severeslump, with the consequent greater stability of their investmentplans, as throughthe actual use of budgetchanges. The recentintroductionof generaladjustmentsof purchasetax as a regulatingdevice is a more promisingdevelopment. It suffers,however,from two defects; the I Reportof the Committeeon the Workingof the MonetarySystem, Cmnd. 827, (1959), p. 183. 2 Ibid., p. 336.

10

ECONOMICA

rFEBRUARY

ratherarbitraryand limitedrangeof goods affected,and the fact that when demandis high there might be expectationsof an increasein purchasetax which would lead to a furtherincreasein demand and when demandis low there might be expectationsof a decreasein tax which would furtherdiminishdemand.Hire purchasecontrols suffer even more severelyfrom these defects. None of the fiscal instrumentsat presentin use satisfiesthe conditions for a satisfactorymeans of correction. If the purchasetax were changedinto a generalsales tax and adjustedby small amounts at frequentintervalsit would do the job. But a preferablealternative would be the introductionof adjustmentsof direct taxes. A fairly simple way of adjustingdirect taxes would be to calculateevery tax to be paid in exactlythe sameway as is now done and then to add or subtracta certainpercentageto the calculatedfigureas a stabilisation adjustment.In the case of P.A.Y.E., which is calculatedon the basis of income and allowancescumulatedfrom the beginningof the tax year,the figureaddedor subtractedin eachpay periodas a stabilisation adjustmentwould be kept separatefrom the figurefor the normaltax and wouldbe neglectedin formingthe cumulatedtax paid. The figure calculatedfor the normaltax in each pay period would thus not be affectedby earlierstabilisationadjustments,and no changewould be needed in existingP.A.Y.E. tables or in the method of using them. The percentageto be added or subtractedcould be changedif necessary at regular intervals; qaarterlyintervals might prove to be shortenough,thoughchangesat monthlyintervalsshouldbe possible. In this case those P.A.Y.E. taxes which are paid quarterlyshould be adjustedby the averageof the percentagesprevailingin the preceeding three months. Similarlyall annual assessmentsfor direct tax should be calculatedas at present and then the average of the percentage adjustmentsprevailingoverthe yearappliedto this figure. Sufficiently fine adjustmentwoald probably be obtained if the changes in the percentageto be added or subtractedwere made in steps of 2-. per cent.,i.e., sixpencein the pound. The absolutevalueof the tax adjustmentwould of coursebe muchlargerfor a personwith a high income than for one with a low income. This has advantagesfrom the point of view both of equityand of efficacy; the largeradjustmentis needed to inducethe personwith the higherincometo changehis expenditure. difficultiesin introducing Therewouldno doubtbe someadministrative a schemeof this sort, but once it was in operationthe additionalwork involvedwouldnot be verygreat. I believethat the choicelies between acceptingthe minor inconvenienceof such a scheme and accepting the continuationof the fluctuationsin employmentand economic activitywhich we have experiencedsince the war. 4.

EMPLOYMENTAND INFLATION

I have so far been discussingfluctuationsin employmentand economic activitywithout any referenceto the averagelevel about which

1962]

EMPLOYMENT,INFLATIONAND GROWTH

1I1

employmentis fluctuating. Considerationof the average level of employmentbringsus to the questionof the relationsbetweenemployment, or unemployment,and inflationand the rate of growth. In the past few years a numberof people have carriedout empiricalstudies of the relation betweenunemployment,or some other index of the demandfor labour,and the rate of changeof wage ratesor earnings. Somewhat different methods and hypotheses have been used by differentpeople. In my own verycrudeattemptto studythis relation' I assumedthat changesin the cost of living only affectwage changes in years when prices are risingrapidly,usuallyas a result of rapidly rising import prices. Others2have assumedthat changesin the cost of living have a proportionateeffecton wage ratesin everyyear. For the post-waryearsthis is probablynearerthe truththanthe assumption I used. But we must then recognisethat changesin the cost of living are in turn mainlythe resultof earlierchangesin wage rates and to a lesser extent of changes in import prices. If these two behaviour relationsare fittedto empiricaldata we can proceedto eliminateprice changesand obtaina singlerelationexpressingwage changesin terms of unemploymentand changes in import prices, or alternativelywe may eliminatewage changes and expressprice changes in terms of unemploymentand changes in import prices. These new relations are not, of course,behaviourrelations,but they are valid relationsfor predictionpurposes,and areindeedin the most usefulform for prediction. The relationI obtainedis best consideredas a predictionrelation of this sort. If the other studiesare also interpretedin this way there is reasonableagreementin the resultsobtained. It seems that if the averagelevel of unemploymentwere kept at a little less than 2-1per cent. the averagerate of increasein wages over a period of years could be expectedto be about 2 per cent. per annumso that with the rate of increaseof productivityexperiencedsince the war the average level of priceswould be almost constant. Also, in the range between 1-1per cent. and 2-- per cent. unemployment,for every 0. 1 per cent. that the averagelevel of unemploymentwas reduced,wages and prices wouldrise at about 6. 3 per cent. per yearfaster. If it is true that such a relationholds we are facedwith a difficultchoice. Thenwe can only reduceinflation,for any given rate of increaseof productivity,at the cost of higherunemployment.I think such a relationdoes hold now, and unless it can be changedwe shall probably move towards a compromisesolution with a ratherhigheraveragelevel of unemployment than in the past few yearsand a lower, thoughnot zero, speed 1 " The RelationBetweenUnemploymentand the Rate of Changeof Money Wage Rates in the United Kingdom, 1861-1957", Economica,vol. xxv (N.S.), (1958). 2 See L. A. Dicks-Mireauxand J. C. R. Dow, " The Determinantsof Wage Inflation: United Kingdom, 1946-56 ", Journal of the Royal Statistical Society, vol. 122 (1959); L. R. Klein and R. J. Ball, " Some Econometricsof the Determination of Absolute Prices and Wages ", EconomicJournal,vol. LXIX (1959), and R. G. Lipsey," The Relation BetweenUnemploymentand the Rate of Changeof Money Wage Rates in the United Kingdom, 1862-1957: A FurtherAnalysis", Economica, vol. xxvii (N.S.), (1960).

12

ECONOMICA

[FBRUARY

of inflation; perhaps about 2 per cent. unemploymentwith about 1 per cent. per yearrise in prices. To consider whetherthe relation can be modifiedwe must know why it is that wagescontinueto risewhilethereis significantunemployment. A numberof possiblecausesare often mentioned,in particular lack of mobilityof labourandindustry,resultingin unevengeographical and occupationaldistributionof unemployment,competitivebidding by employersfor the most suitablelabour and-trade union pressure. The mobility of labour and industrywould be increasedif geographicaland occupationalmovementsin relativewages were allowed to take place more freely. If wages in areas and occupationswhere unemploymentis low or excess demandis high rise more than those whereunemploymentis high or excess demandis low, therewill be a greaterincentivefor labour to move to the areas and occupationsin which wages have risen most, and for industryto move to the areas wherewageshaverisenleast, and also for industryto adaptits production methodsto use more labourin those occupationsin whichwages have risen least. Competitivebidding up of wages in a particularoccupation and area by employers can easily occur even when some labour in the same occupationand area is unemployed. The basic reason for this is the wide range of abilitythat exists amongdifferentindividuals in the same occupation. Somepeople,even in a singlenarrowoccupational classification,are worth to an employer considerablymore than the averagerate of pay in the occupation; othersthroughindividual defects of character,intelligenceor physique,are worth less. A wide awake employerwill often find it profitableto pay 5 or 10 per cent. above the averagerate of wages for a particularclass of labour in his locality. In this way he can choose the best men and may well finishup with employeeswhoseproductivityis 10 or 15 per cent. above the average. If a large proportionof employersadopt this practice wages may be bid up quite rapidly even when there is a significant amount of unemployment. The best solution would seem to be to allow more flexibilityin the wages paid to differentindividualsin the same occupation. The fourth possiblereason I mentionedfor wage rates rising when there is significantunemploymentwas the power and pressure of tradeunions. I have some doubtswhetherthis has been an important factor,but if it has, and if the tradeunionsfully understandthe results of their actions, it can only be counteredat the cost of occasional major strikes. But have the results of an irresponsibleuse of their powerbeen made clear to trade unions by Governments? If it were widely and clearly understoodamong the membersof trade unions that the full use of their power to force up money wages would only lead, at a given level of unemployment,to a faster rate of inflation and that the Governmentwould have no alternativebut to check this higherrate of inflation,in part at least, by loweringdemandand

1962]

EMPLOYMENT,INFLATIONAND GROWTH

13

causing some increasein unemployment,it seems possible that the tradeunionsmight see wheretheir trueinterestslay. I am not so naive as to expectthat the suggestionsI have made for trying to shift the relation betweenemploymentand the rate of rise of wage rates in a way which would make it possible to maintaina higherlevel of employmentwith anygiven speed of inflationarelikely to meet with an enthusiasticresponse. They are all suggestionsfor moreflexiblearrangements whichwould allow freerplay to the market forces of supply and demand. But for at least a centuryand a half beforethe secondworldwar wage-earnersfrequentlyhad a prettyraw deal from the marketforces of supply and demand,for in most of the yearsof the tradecycle,and especiallyin the catastrophicinter-war years, there was a deficiencyof aggregatedemand. The traditions built up over that period still persist and will continueto persistfor some yearsyet. But in due courseit may be realisedthat continuously rising standardsof living come only from continuouslyrising productivity and that provided the Governmentis not prevented by inflationfrommaintaininga high aggregatelevel of demandfor labour the marketforcesareno longerharmfulbut play a vital partin helping adaptationand progress. 5.

EMPLOYMENT AND GROWTH

I have suggestedthat a slightlyhigher averagelevel of unemploymentthanthat whichwe havehad in the last few years,perhapsa little over2 per cent.,may be acceptedas a necessaryconditionfor moderating the speed of inflation. One often hearsheated argumentsagainst checkinginflationin this way, on the groundsthat if the rateof growth of the economy were more rapid prices would not rise so fast, and that the rate of growthwill be reducedby operatingthe economy at a slightlylowerlevel of employment.It is true,of course,that a higher rate of growthwith the samerate of changeof wages would lead to a lower speed of inflation. It is also true that while unemploymentwas actuallyincreasing,let us say from 1 per cent. to 2 per cent., output would be risingless rapidlythan it would have been if unemployment had been kept at 1I per cent. But the argumentis often phrasedas if the steadyrate of growthof the economywith unemploymentconstant at 2 per cent. would be less than the steadyrate of growthwith unemploymentconstantat 12 per cent. I doubt whetherthis is true. During short-periodcyclicalfluctuationsthe variationsin outputas a percentageof the growthtrend are about five times as large as the variationsin the percentageunemployment.The differenceis largely accounted for by variationsin short-timeand overtime; by some people, mainly marriedwomen, moving into and out of the labour force; and by " hoarding" of labour during a recession which is expectedto be short. If unemploymentwere to rise by 2 per cent. and stay at the new level, the hoardingof labour and some of the initial change in short-timeand overtimewould be only temporary

14

ECONOMICA

[FEBRUARY

and when the adjustmentswere completed the percentagefall in output would be considerablyless than five times the increasein ihe percentageunemployment.If accountweretakenof the valueof leisure the fall in real incomewould be still smaller,sincethat part of the fall in outputwhichwas due to a reductionin overtimeand movementof people out of the labour force would be partly compensatedby the value to the people concernedof the extra leisure. If an increaseof i per cent. in unemploymentcauseda decreaseof 12 per cent. in output, with equal proportionaldecreasesin investmentand non-investment expenditures,investmentwould fall by 1 per cent. of its own value, and assumingthat the growthrate was proportionalto investment the growthrate would also decreaseby 1I per cent. of its own value,for example,from2.5 percent.to 2.4625 percent.; an extremely small decrease. If all the decreasein outputresultedfrom a decrease in non-investmentexpenditurethere would be no change in the rate of growth.The main conclusionfrom this is that the differencein the steadystateratesof growthbeforeand afterthe transitionto the higher unemploymentwould be extremelysmall if the growth rates were proportionalto investment. Other influences,such as possibleextra incentiveor compulsionto invest in cost-reducingequipmentmight easilyoutweighany small differencedue to the slight changein investment. It is sometimesargued that the decreasein demand would slow down growthby reducingthe desireof firmsto invest; but this could always be remediedby reducinginterestrates. Indeed since we are assumingthat a Governmentcould hold unemploymentat the new level, investmentin the new steadystate, as in the earlierone, would have to be broughtinto equalitywith savings. It is on the willingness to save, and the more generalinfluencesof educationalimprovement, research,and so on, that the rate of growthdependsin presentcircumstances. I do not think that very small changesin aggregatedemand and unemploymenthave much effecton these. 6.

RATES OF EXCHANGE

The final questionI wish to discuss,very briefly,is that of rates of exchange.The othermajortradingcountrieshaveproblemsof employment,inflationand growthwhich are similarto those of Britain. The questionariseswhetherall countriesare likely to hold that balanceof internalobjectiveswhichwould be consistentwith the maintenanceof fixed exchangerates. Professors Samuelson and Solowl have considered the relation betweenunemploymentand the rate of changeof the consumerprice indexin the United States. Theirtentativeconclusionis that assuming continuationof the conditionsof the post-warperiod the price index might be stable if unemploymentwere held at 5 to 6 per cent., and I P. A. Samuel.on and R. M. Solow, " Analytical Aspects of Anti-Inflation Policy ", AmericanEconomicReview,vol. L, no. 2 (1960).

1962]

EMPLOYMENT,INFLATIONAND GROWTH

15

might rise at about 2 per cent. per annum if unemploymentwere 4 per cent. Some estimateswhichI have made lead me to think that the situationin the UnitedStatesis less favourablethanthis. I estimate that 7 to 8 per cent. unemploymentwould be needed to maintaina stable price level, and that at 4 per cent. unemploymentthe price level would rise at about 4 per cent. per annum. Of course efforts to increasethe rate of growthand to reducestructuralunemployment may improvethe relationbetweenunemploymentand price changes; but unless my estimatesare badly out or considerableimprovements are obtainedit seemslikely that if unemploymentis reduced,as seems to be hoped, to 4 or 5 per cent., the United States may well have a rather faster rate of inflation than Britain would have with 2 per cent. unemployment. In Germanysimilarproblemsare beginningto appear. Unemployment has fallen steadilyfrom about 10 per cent. in 1950 to 1.2 per cent. in 1960, and labour costs are now rising rapidly. The growth rate is still high but some specialfactorswhichcontributedto it, such as reconstructionafter the war and the currency difficulties,and availabilityof skilled labour from unemployedor from refugees,are passing. It may well be that in Germanytoo the problemof choosing betweeninflationand unemploymentwill become acute. These three countries,and others, may perhapsbe preparedto adjust their own choices about internalbalancein orderto make them consistentwith a regimeof fixed exchanges. But I have some doubts whetherthey will be preparedto do so, or if they are whetherthey in fact know the quantitativeworkingsof their own and other economieswell enough to choose the appropriateobjectivesbefore gradual divergenciesin pricelevelshave cumulatedto such a degreethat they imposea heavy strain on the internationalmonetarysystem. If this is so, it might be betterto allow a limitedflexibilityinto the exchangesystem,so that gradualdriftsin relativeprices,whichwould probablynot be at a rateof morethan 1 or at most 2 percent. peryear, would not producecumulativedisequilibriain balancesof tradewhich might eventuallynecessitatelarge and suddenmovementsof exchange rates and do seriousdamageto the internationalmonetarysystem. A limited flexibilitycould be introducedby an agreementthat the par value of any one currency,in termsof gold, could be changedat any time providedthe total changein one directionin any periodof twelve monthsdid not exceed1 percent. Therewould,I think,be muchmore confidencethat a countrycould in fact work withinthis rule than that it could for ever succeed in keeping the par value constant, so the fear,or hope, of a suddenlargechangein the rate,with the tremendous speculativemovementsit causes, would be greatly diminished. The maximumpermissiblerate of change of the par rate, 1 per cent. per year, could easily be offset by short-terminterestdifferentials,so it neednot lead to any majortransfersof capital. And this limiteddegree of exchangeflexibilitywould allow each countrytime to find by trial

16

ECONOMCA

[FERUARY

and errorthat compromisebetweenits internalobjectiveswhich was consistentwith its exchangerate policy. I said at the beginningof this lecturethat I believedone of the main difficultiesin devisingand implementingappropriateeconomicpolicies is lack of quantitativeknowledgeand understandingof how the economic systemworks. By now I have no doubt amplydemonstratedat any rate my own lack of knowledgeand understanding,and it only remainsto apologiseto any of you who may have come hereexpecting clear and definiteanswersto the problemsI have been discussing. I hope the next personto face the ordealof givingan inaugurallecture on electionto the Tooke Chairwill be in a position to explainthese mattersmore clearly. The London School of Economics.