18 The Phillips Relation The previous two chapters have considered two contrasting hypotheses about price adjustment. In neoclassical analysis prices adjust fully to bring about market-clearing equilibrium. A distinctive feature of the neoclassical model is that the equilibrium values of the real variables are invariant with respect to changes in the equilibrium value of the price level brought about by changes in stock of money. In Keynesian analysis prices either fail to adjust at all or only do so slowly so that markets remain uncleared. So from both perspectives interest focuses on how the price level varies over the trade cycle and how such variations relate, if at all, to fluctuations in real output and employment. 18.1 The early Phillips curve Attention was drawn to a statistical relationship between unemployment and inflation by Phillips [1] in a now classic study published in 1958. The form of the relationship studied by Phillips was between the annual rate of change of money wages and the annual average percentage rate of unemployment using UK data for 1861-1957. Phillips concluded from the data that the rate of change of money wages seems to be inversely and non-linearly related to the percentage rate of unemployment. He displayed this relationship by fitting a curve to the data. The original Phillips curve is reproduced in Figure 18.1. Because the observations for 1948-57 lie quite close to the curve fitted for the years 1861-1913, it was thought that the relationship was a stable one that would persist over a long period of time. One can move from a relationship between the rate of change of money wages and unemployment to one between the rate of change of the price level and unemployment by allowing for long-run changes in the productivity of labour. If the money wage rate increases at the same rate as labour productivity, then the average labour cost of producing goods remains unchanged. If goods prices are some stable mark-up on average costs, then prices will not change so long as the rate of increase of money wages equals that of labour productivity. One can therefore subtract the average rate of 338The Phillips Relation 339 gn BO | os 3 3 Curve fited to 1861-1913 date Fe] \y ga os eas BE 4 Be 3 23 B54 s 0 3 °o z Unemployment (per cent) FIGURE 18.1 The original Phillips curve increase of labour productivity, Q/Q, from the rate of increase in money wages, W/W, to obtain the resulting rate of increase in the price level, P/P." That is PW @ (18.1) P WO In the 1960s Phillips's work generated many more empirical studies of the relationship between the rate of change of money wages, inflation and unemployment. Other variables, not only unemployment, were tried as explanatory variables for wage or price inflation. Some studies found unemployment to be insignificant in explaining wage inflation, though these results tended to be for the post-war years using annual data. Studies using quarterly data were more successful in securing a significant effect for unemployment. On the whole the studies carried out in the 1960s tended to show a significant non-linear relation between wage inflation and unemploy- ment. (A few of the early studies are summarised in Table 18.1, p. 354.) Not much attention was paid at the time to evidence which suggested that the Phillips curve was not stable over time. Two influential early studies (Phillips’s own and Lipsey’s [2]) did find that the relationship shifted over time as indicated by differences in the estimated coefficients for different periods. Q a1 Qa Q Ww dwt Woda Ww340 Current Controversies in Macroeconomics Policy inferences Phillips inferred from his estimates that with 5+ per cent unemployment the aggregate money wage rate would be stable. Given that labour productivity grew on average by about 2 per cent a year and that the relationship between W/W and unemployment is non-linear, he concluded that maintaining a stable price level would require about 24 per cent unemployment. Other economists (notably Samuelson and Solow [3]) proceeded to treat the Phillips curve as offering policy-makers a menu of choice between inflation and unemployment. On the assumption that the Phillips curve is stable, a permanently lower level of unemployment could be achieved at the cost of a higher rate of inflation. It was up to policy-makers and the electorate to choose the point along the Phillips curve which gave the preferred trade-off between unemployment and inflation. The notion that a higher rate of inflation can give a permanently lower level of unemployment is at odds with the neoclassical macroeconomic model. In that model, as shown in Figure 6.3 (on p. 85), the level of real output and hence unemployment is invariant with respect to the price level. An increase in the price level, achieved by shifting the aggregate demand function up to the right, has no effect on long-run real output. On the other hand, a stable Phillips curve does provide a dynamic analogue to the analysis of the effects of changes in aggregate demand in the Keynesian synthesis model. Here an increase in aggregate demand raises the price level relative to the money wage rate and this causes a permanent expansion in employment and output (see Figure 6.5, p.89). In the comparative-static world of this model output attains a higher equilibrium level with a lower real wage rate. At this new equilibrium the price level has moved up to a new constant level while money wage rates remain unchanged. Theoretical underpinnings: excess demand The Phillips curve started life as an empirical observation: its theoretical foundations were rather sketchy. Phillips's own explanation, which was elaborated by Lipsey, is that given a stable rate of change of labour productivity and the absence of sizeable import price fluctuations, money wages rise more rapidly the greater the amount of excess demand in the labour market. Formally, this is expressed as (18.2) where W/W= (1/W)(dW/d0) is the rate of change of the money wage rate, L” is the demand for labour and L° is the supply of labour. Now the demand for labour consists of those in employment, E, plus vacancies, V. The supply of labour is equal to employed workers plus the unemployed, U. Therefore ges) (1833) Ww E+uThe Phillips Relation 341 Since vacancies vary inversely with the unemployment rate and did so in a stable way up to the mid-1960s, unemployment by itself could be used as a measure of excess demand in the labour market. So we arrive at the Phillips relation a) (Sr W) <0) (18.4) dU Equation 18.4 only specifies an adjustment mechanism. It tells us nothing about whether the disequilibrium which sets off the rate of change in money wages is caused by factors from the demand, or the supply side, or both sides of the labour market. The coexistence of positive unemployment with rising money wage rates can be explained by the existence of frictional unemployment. Even when the labour market is in equilibrium, so that the demand and supply of labour are equal and there is no tendency for the money wage rate to rise, some frictional unemployment will exist. This would be the level of unemployment at which the Phillips curve cuts the horizontal axis. Any reduction in unemployment below this implies excess demand for labour and results in rising money wages even though unemployment is still positive. ‘An alternative but not mutually exclusive explanation for the Phillips curve cutting the horizontal axis at a positive unemployment rate is the aggregate nature of the statistical relationship. It is derived from aggregating over many individual labour markets. In each labour market the money wage rises when there is excess demand and falls when there is excess supply. In the Keynesian tradition it is supposed that the wage rate rises more rapidly for a given value of excess demand than it falls when there is equivalent amount of excess supply. The wage dynamics of each labour market would then be depicted as in Figure 18.2. ‘At any one time some markets will experience excess supply, others excess demand. When one averages over all markets then a positive average rate of unemployment due to the excess supply markets can be accompanied by a ing average money wage rate because wages in the excess demand markets rise more rapidly than they decline in the excess supply markets. The higher the W | Rate of change WW | of money wage rate Weak —_Net excess demand for labour bs3 Le Fiure 18.2. Wage adjustment in an individual labour market342 Current Controversies in Macroeconomics general level of aggregate demand, the lower the average aggregate unemploy ment rate and consequently the faster is the increase in the average money wage rate. (A further exposition is given in Barro and Grossman [4].) In the light of recent neo-Keynesian work one can see that this kind of theoretical rationalisation of the Phillips curve relies on the supposition that markets fail to clear. Money wages are changing because net excess demands in labour markets are not zero. A permanently lower level of unemployment can be achieved at the cost of higher inflation by maintaining a continual state of excess demand in the labour market. This can only occur if there is no adjustment to bring about a cleared labour market. The mechanism by which a more rapid rate of inflation could secure a permanently greater supply of real output was not at first properly investigated. Nevertheless, the belief that infla- tion could permanently lower the level of unemployment spread far beyond the economics profession in the 1960s and was generally regarded as a viable political option 18.2. The natural rate of unemployment hypothesis Cynics have observed that whenever an apparently stable economic relationship is discovered and then used for policy purposes, it promptly breaks down! The most famous example of this is probably the Phillips curve. The relationship between inflation and unemployment in the United Kingdom from 1960 to 1981 is shown in Figure 18.3. This shows pretty conclusively that the 5 x Phillips curve for 1861-1913 xa Percentage rate of change of the retail price level o 7 2 3 4 5 6-7 8 38 1 Unemployment {percentage rate) FIGURE 18.3 The Phillips relation in the United Kingdom, 1960-81 Note: data up to 1981 are up to the second quarter Source: Economic Trends.The Phillips Relation 343 Phillips curve has not been a stable relationship; over the last fifteen years inflation and unemployment have both shown a marked secular increase. Other industrialised economies have displayed a similar tendency, though the levels of inflation and unemployment of some of these economies have remained below British rates. (See Figure 1.1 on page 4.) The breakdown of the empirical Phillips relationship in the late 1960s coincided with new theoretical work, notably by Friedman [5] and Phelps [6], which denied the existence of a permanent trade-off between inflation and unemployment. The starting-point of this analysis is that microeconomic theory posits a relationship between the level of excess labour demand and the rate of change of real wages, not money wages as suggested by Phillips. Instead, the Phillips relation should be wo TIO (18.5) where w/w = (1/w\(dw/dt) = rate of increase of the real wage rate. By defini tion, the actual rate of change of real wages equals the rate of change of money wages, W/W, minus the rate of inflation, P/P: Ww P * (18.6) wo WP . ‘When workers and employers set the money wage rate, each party is really concerned with the real wage rate at which labour will be hired. The perceived real wage rate implied by a particular money wage rate depends on what is the expected rate of inflation. Unless the economy is in long-run equilibrium so that expectations are always realised, there will be some divergence between the expected rate of inflation and the actual rate of inflation. This means that the appropriate relationship between the rate of change of the real wage rate anticipated by a worker and the rate of change in the money wage rate is E (3) (18.7) where E(P/P) = the expected rate of inflation. If workers are rational, they fully adjust the increase in money wages for the expected increase in prices to obtain the resulting change in the real wage rate upon which they base their decision as to whether to remain in their present employment or to continue job search if unemployed, as the case may be. If workers do not fully take into account the inflation that they expect to occur when estimating their real income from their money income, they are said to have ‘money illusion’. Behaviour due to money illusion is quite distinct from incorrect expectations. One can overestimate one’s future real income because a higher rate of inflation occurs than one expected. This is distinct from money illusion, which would cause one not to act in response to one’s future real income, even though one is correctly anticipating the rate of inflation that will344 Current Controversies in Macroeconomics occur. A coefficient, a, is subsequently attached to the price expectations vari- able. It will equal 1.0 if employees are both rational (that is, they do not suffer from money illusion) and can adjust fully their money wages to compensate for expected price increases. If from equation 18.5 we substitute /(U) for w/w in equation 18.7 and rearrange terms we get ” pus or (2 (18.8) U) + aE (— . i (7) where a = 1 if, given the rate of unemployment, workers completely adjust their money wage to compensate for expected inflation, and 0< a < 1 if they only partially adjust their money wage. Equation 18.8 is known as the expectations-augmented Phillips curve. The original Phillips relationship, shown in Figure 18.1, did not contain the a£(P/P) term since it was based on the implicit assumption of a zero expected rate of inflation. When the rate of inflation was very low such an assumption was reasonably plausible. Once a positive rate of inflation becomes anticipated and given that a # 0, then the rate of increase of money wage rates at all levels of employment will adjust to reflect anticipations about inflation. In terms of the Phillips-curve diagram, this means an upward shift in the entire relationship. There is therefore a whole family of short-run Phillips curves, each one corresponding to a given rate of expected inflation. Friedman [5] proceeded to argue that the expectations-augmented Phillips curve would shift in such a way that in the long run a higher rate of inflation would result in no change in unemployment. This argument is explained in Figure 18.4 and it assumes that a — 1 in equation 18.8. We start with an economy which has a stable price level and constant real and money wages. We are simplifying by assuming no growth in labour productivity. The short. run Phillips curve for a zero rate of expected inflation is PCs, Since the price level is stable the unemployment level is Uy. This level of unemployment is termed the natural rate of unemployment, It is defined as that rate of unemployment which is consistent with labour-market equilibrium and at | Long-tun Phillips curve PC IEVP/P=0) 1 PC\(E(P/P) = 0.05) 0.05} - - Rate of change of money wages 0 Unemployment FIGURE 18.4 The expectations-augmented Phillips curveThe Phillips Relation 345 which the price level could be stable. The natural rate of unemployment is determined by the real factors which affect the amount of frictional and structural unemployment in the economy. The government has been told that there is a trade-off between unemploy- ment and inflation. It therefore chooses to keep the economy at point A of the short-run Phillips curve PC, by expansionary policies which increase the money supply. The rate of inflation now rises to 5 per cent and the level of unemployment falls to U,. We need an explanation of why unemployment falls and output rises when the rate of inflation increases. In the neoclassical interpretation of the Phillips relationship this occurs only because the inflation is unanticipated. Since demand has increased firms start raising prices and bidding up the money wage rate to attract more labour. Because workers’ expectations of inflation are below the actual rate of inflation, they think that the higher money wages now being offered means that real wages have risen. The supply of labour therefore increases. This is shown in Figure 18.5 as a downward shift in the labour supply function from SS to S'S". When actual and expected inflation are equal the labour supply schedule is SS. When infla- tion increases but expected inflation lags behind workers are deceived into offering to work for a lower real wage. The demand for labour increases as firms move down the demand for labour schedule. Unemployment falls and output rises as the economy moves up the short-run Phillips curve PC. As expectations adjust towards the actual rate of inflation, workers realise that real wages are lower than they had anticipated and therefore require a more rapid increase in the money wage rate. The supply of labour schedule shifts back up until it regains its initial long-run position once expected and actual inflation are equal. As the supply of labour shifts back to its original position, the short-run Phillips curve also shifts outwards because the expected Tate of inflation is rising. When expectations have fully adjusted to the new igher rate of inflation the short-run Phillips curve in Figure 18.4 has shifted up to PC,, which is its position when the expected rate of inflation is 5 per cent. The economy is now at point B in Figure 18.4. Unemployment is back to its natural rate but there is now a 5 per cent rate of inflation. (This requires a per manently higher rate of increase in the money supply.) The idea that there is no oe ctea 2. om 5° monet sacl 6 expected acral erpected Bescwat (This lagtam imple that the actual fei nasa ols ehenenplonmen (sree This eed ot be te eee iit tan marl eos ae productivity is rising) e Real wage rate 0 tty ‘Quantity of labour FIGURE 18.5 The effect of differences between the actual and expected rates of infla tion on the supply of labour346 Current Controversies in Macroeconomics way in which the rate of unemployment can be permanently held at a different level to the natural rate of unemployment is known as the natural-rate hypothesis (NRH). Another way of stating the same point is that the long-run illips curve is vertical. s also implies that if the government attempts to bring down the rate of inflation, unemployment will temporarily rise above the natural rate if expected inflation adjusts with a lag. For a while the expected rate of inflation based on past experience will lie above the actual inflation rate. Workers are offered a slower rate of increase in money wages by employers, given their increasing inability to raise prices in product markets at the previously experienced rate. As workers’ expectations of inflation have not adjusted downward with this new development in the product market, they bargain for money wage rates in line with their unadjusted price expectations and their reservation money wage rate rises in line with these expectations. This leads to an upward shift in the supply of labour schedule in Figure 18.5. Real wages rise and employment falls. At first the economy moves down the current short-run Phillips curve (such as PC, in Figure 18.4) but as expected inflation falls the short-run Phillips curve shifts downwards and the natural rate of unemploy- ment is restored once actual and expected rates of inflation are equal. The Keynesian counter-argument The natural-rate hypothesis, as we have just seen, depends on the expectations- adjustment coefficient, a, being equal to 1.0. Keynesian economists (see Tobin {7}), while accepting that a positive rate of expected inflation will reduce the size of the trade-off, have still maintained that because a is less than 1.0 some long-run trade-off occurs even when expected inflation has fully adjusted to actual inflation. To see why a long-run trade-off between inflation and unemployment will exist when 0 0. In conclusion, it has been shown that under adaptive expectations: (a) there is no long-run trade-off between unemployment and a steady rate of inflation; and (b) a short-run trade-off exists as unemployment falls when inflation accelerates. Point (b) offers the possibility of permanently lowering unemployment by con- tinually accelerating inflation. This is known as the acceleration hypothesis. 18.5 Rational expectations A powerful criticism of the adaptive-expectations hypothesis is that it assumes people keep basing their expectations on the values of lagged variables and fail to learn from their past errors. When inflation is rising the error between actual and predicted inflation is positive over successive periods, and so is serially correlated. This error is repeated and so is called systematic, yet no attention is paid to it if expectations are adaptive. If economic agents are rational, they will make full use of all the available information when forming expectations and not just rely on past values of the relevant variable. Expectations which are conditioned on all the available infor- mation are called rational expectations. When expectations are formed rationally the errors between actual inflation and the expected rate are random and are serially uncorrelated or independent over time. There is no systematic error by definition because any systematic error would reveal that information had not been fully used when the expectations were formed. So the expected or mean error is zero. If expectations are rational, then (5)-#(6)s ue where ¢, is a serially uncorrelated error term with a zero mean. When expectations are formed rationally then people use information derived from the model which they think explains how the economy behaves. This means that if inflation is due to monetary expansion, then informationThe Phillips Relation 353 about the current movements in the money supply will be important in con- ditioning expectations. For instance, if the rate of inflation is determined by Py_,(™ (18.21) ()=° (ar),*® where (M/M) is the rate of growth of the money stock, then ®(5) ° (a7), (18.22) Substituting 18.21 and 18.22 into equation 18.15 (p. 351) gives ol (F)- (3) +e (18.23) The application of rational expectations therefore leads to the conclusion that there is not even a short-run trade-off between inflation and unemployment. The short-run Phillips curve in Figure 18.4 shifts up to PC, instantly. Equation 18.23 shows that unemployment is affected only by random errors, i.e. by unpredictable events. The government can only secure a short-run decrease in unemployment if it makes surprise increases in the money supply. Although the instantaneous rate of adjustment derived from the full information application of rational expectations may seem unrealistic, its basis, that people will make use of all the available information when forming expectations and not make correctable errors, is quite sound, SG, 18.6 Empirical evidence The numerous empirical studies of the Phillips relation have been concerned with finding out: 1. Whether there is a short-run trade-off between inflation and unemploy- ment. This requires a negative and significant relationship between wage or price inflation and unemployment. 2. Whether there is a long-run trade-off. It has often been presumed that the absence of a long-run trade-off depends on a, the price expectations coefficient, being equal to 1.0. 1. A short-run trade-off? ‘The evidence for this is mixed. Of the studies summarised in Table 18.1, the following found unemployment or its proxies to be significant: Dicks-Mireaux (11), Klein and Ball [12], Batchelor and Sheriff [13]. Unemployment is reported as insignificant in Lipsey [2], Johnston and Timbrell [14], Henry et al. {15}, Parkin ([16] and {17]), Coutts e¢ al. [18]. Lucas and Rapping ([19] and (20]) explicitly test the hypothesis that the relationship between unemployment(ere) (ree) (oD eee) 99-9561 EP8'O = Y ‘LO = 8 ‘a/S8O8'0 + 2IEPTOO'D + 91 Z180°0 + SZECO~ onpord— Aqevrenb in, 80860 = 2 (L'0 = palinsse ‘9 ‘siuarayjo0 suoneioadxa oalidepe) yeu 303 xoput zs'e) — (Z90') (IES) (LOE'ED (EZ9'TN) (ZL'0) sou Jo aaUEYO 99-8h61 6961 ad6012'0 + GTLPIT0 — GOESHO'O ~ WE6O'O +9" SLSO + 81900 Jomu=q — jenuUNE AN [TZ] mo}og sooud (900) (P0°0) | a(60°0) «(661 Tuy Jo aBuEYD axvpso~ PI WIe0 + MLO + LET joer ad mak iad pokoydura uosiod sad soureyes (stsz 44 1961 660=.¥ pue‘sfururea —_payeunso) a M280) (00) (ETO) ¥(€9'0) yo adureys 6S-9p6T NRO vittgge-z + 'aor0 + 4460 + 06 goaer= 74 yenuue yn pT +(80°0) «(ZT0'0) — «(00°9) [4s-8v6r) svo= dy ge0=2My 160-14 d69°0+ AVYEO+ »-81 11+ l6e-€z61) S80 = 2a dIZO+AVEVO ¢-9TT + 1-S¥9+ 17 I~ A €161~7981 0961 Tenuue iq [2] dase xeput ‘oud rawinstio> agro (00) (ETOe(10'ETD) Jo af ue Jo HEL 966°0 = .¥ XLEL0'O + —A7WOTZO + (T2A@D)M IZ + S9'SE Ayayrenb = tq ‘saver afm Apjoom atz948 6561 #(0F'0) (2600) *(€10'0) Ayoirenb ut 9s-8rer = (zt) mea £60 = 8 9701 +4706'7 + 4abS8°0 +42160'°0— a8ueyofenuue= 4 Apron YQ pur MOpY (Atnow) saver afem Kuow jo LS61-T98T 8561 (E161~ 1981 Sanwa aBes9Ae 105 PITT) 44-1. M8E9'6 + 006'0- aBuEyD Jo mE= mM —EnUUE XN [1] sdeua suonenbs uowssaidoy ajqetrea 1wapuadeq eq Apmis ffo-apoi uonnyfur—suatudojduaun ayy fo saipnis dusawouors amos fo dapuuns ¥ 1°81 AIAVLys LET 16 loz a-a (05D S190 = Zy0S0"1 HED (S70) ESO) | AMIE), ETO (TO) (760) 466°) TyEezsTO+ ALTO + FLLOTE'O — Z4E91E'0+ LUS9TT-O + Y40LO¥'O + 19Eb9"0 ~ BeRE'E +($95000) albS7T) ave) SLLO=M-a (L900) 74 a a s(.07"D) SL0=.¥ —“ab9SO+ “)exsov0 (2) suet €) B61 — 116 60% =M-G (961) ED). L¥'O) 89°0=.¥ 16697 + 1090+ WILT T + 6 E- MOET OERVO='F — (E070) (860) «(s88"0) 10L'0= 28 MbOTOOPFO=¥ — gd9EVO + FIZO'O + FEOE soo ~ “fis sur sau“ aist “su isu “ASU +(6¢'SD 960 “Ss “Hsu Asus “ISU a 1896 — “As sz (SED s(40'6)—(1'8) 50 su “gg + asm “ype — aise Ze0E = 600e —S ETO ETD A256) (66) (48°98) (69°) (€9'0) eso teO LTT g9e SIT ELT 9st ONE ST os 4, SEU SEEN es, OE wy ig gay awe yay yu tyuy “suc stemq (soo) 070) +(80°0) «(10°0) StOO= eX 'ag0+ !m/MTHO Fa HLT 650 + 7H0'0 sore a8em Aqy90% JO afueyo Jo a1e = m a (s9-9¥61) (sb-0f61) (67-4061) (s9-¥061) 2) qwow -Aoyduraun = 17 aDL96r CD 8y6T Auouenb yn, 99-6¥61 Auouenb 3, 1L~6S561 yenuue yn. (eg D8 ¢6T Ayauenb 0, (uonesuadurcs adem Aouow Apmoy = 7) S9-%061 Tenuue sq, G@noy rad uones suaduros fea ‘Aauout = M4) S9-0€6I reawue sn 9u6t (LI) PEM pur rung, supped 9461 {st} “pw 10 ATW yL6T (et) Tesquity, pur uoysuyos Ler (91) uppea 6961 {oz} auiddey pur ston] 6961 fen) Burddey pur sean]SLO= ee Teuonex 3ut suoneisadxe sunsse pareusnisa 1861 sored ur afuryo lez] (eo) err) (LD> (97-0-)—_parsadxa = ad SL-096T wool pur S90=.¥ 4 RA p80 ~ PA OL — FH(M)BOI CEO — 9900— ad “4 Aueirenb ¥N.—PIOFUT, veo iy suoneioadxa Teuoper Apzed vo +(980) (010°) 4(800'0) "FU pase jonoq 20rd V+4q G01 v 1198'0 + 49 FOL 947-0 — ! 44/301 9PI0'0+ BTID Parada = !+4y anon se = 14y. sfurusea 4qsndq 9L6t (E80) +(S10°0) +(€10°0) «(SZ0'0) —aBts2a8 Jo aunyo oL~zs61 lec] 1434 31 v gg08'0 + 17,7 Bor seo'0 ~ !~*(u/) FOI EITO'O + 689° jooer=*4 = Ayayenb $y wnt eEDoW suopeioadxa Teuorrex wo pposeg Joao] a0t1d 190-4 paroadxa = Fy sfumusea