Rational and non-rational expectations of inflation in Korea

WC&I De~lopmetrr. Vol. Printed in Great Britain.

Rational

16. No. I. pp. 195-i-205, 198X.

0305-750Wsx 53.W + o.llo 0 I9SS Prrgamon Press plc

and Non-rational Expectations Inflation in Korea

of

KYE SIK LEE* Korea Development Institute, Seoul Summary. -A modest attempt has been made to investiyate the expectations hypothesis for the Korean economy since 1965 within the framework of a simultaneous equations model using a variety of expectations formation schemes. The expectations models explored in the study are of weighted rational and non-rational expectations, which include static. adaptive, extrapolative, and normal expectations. The empirical evidence reveals the following: (i) the expectations hypothesis cannot be supported, implying that for the Korean economy. wages do not fully adjust to inflationary expectations; (ii) inflationary expectations play an essential role in money-wage determination in the non-rational models, but not in the rational models; and (iii) the normal expectations model best portrays the role of inflationary expectations in the economy.

two categories: rational and non-rational expectations. The investigation of the expectations hypothesis involves simultaneous testing of two hypothesis: that price expectations play an essential role in the determination of wages, and that these expectational variables enter the wage equation with a coefficient equal to unity. Sometimes the first hypothesis is called a weak version of the expectations hypothesis, while the two hypotheses as a whole are referred to as a strong version of the expectations hypothesis. The plan of the study is as follows. The empirical model is presented in Section 2. Section 3 explores various hypotheses of expectations formation. Section 4 then discusses the estimation results obtained. Finally, Section 5 concludes the paper.

1. INTRODUCTION Over the last decade increasing attention has been devoted to the role and behavior of inflationary expectations, and particular concern has been directed to testing the “expectations hypothesis.” The hypothesis posits that wages and prices fully adjust to inflationary expectations. While this hypothesis has given rise to a substantial body of empirical testing and analysis, most studies are open to criticism on at least one of the following two accounts. One is that most of the work has focused on the estimation of a single wage equation and thereby failed to incorporate the simultaneous determination of crucial variables in the model. The other is that they lack a sufficiently general treatment of the expectations formation schemes, given that estimation results can be drastically different, depending on the proxy variables one uses for the unobservable price expectations variable. The principal concern of the present study is to examine the expectations hypothesis for the Korean economy since 1965 within the framework of a simultaneous equations model using a variety of expectations formation schemes. The model comprises three equations, incorporating the simultaneity of wage, income and price determination. Given that a single approach of inflation expectations cannot be viewed as an adequate representation of a complex process. a number of alternative hypotheses have been formulated and explored which fall broadly into

2. THE MODEL The

model

following

to

three

be

estimated

is composed

of the

equations:

*Gratefully acknowledged are the helpful comments and suggestions from research fellows at KDI and from the seminar participants at the National Taiwan University and the Institute of Developing Economies, Tokyo, Japan. I wish to thank an anonymous referee for valuable comments. I am also indebted to Yongsoo Kwon and Hongsik Min for most capable research assistance. Remaining errors are solely mine.

195

WORLD

1%

Ii,. (32% 13,> 0. Ii: < 0

‘1,.

‘{:

>

0.

‘{2 <

I)

DEL’ELOPXlEIU-I

(2) (3)

where 1Y = nominal wage. 11-l = inverse of unemployment rate, J = real GNP, P = actual inflation. P’ = expected inflation. 111 = nominal money stock. G = nominal government spendin g. c = real consumption, X = labor productivity.. and P” = import price. Equation (1) embodies an expectationsPhillips relationship. The first augmented variable. II is a proxy for excess supply of labor. The second variable, _v serves as a proxy for the demand for labor. For a period of rising business activity, with the demand for labor increasing. firms are likely to make bids and offers more actively for workers than otherwise. For a period ot falling business activity, just the opposite will be the case. The demand conditions for labor can be reasonably reflected by the variable J since it will be the most a,eneral indicator of changing business activity.’ The third variable. P is supposed to capture the increases in wages caused by increases in prices through the equalization of the wage rate to the margtnal revenue product of labor. The role of actual inflation in addition to expected prices in the determination of wage has been emphasized by some other studies. such as Kuh (1967). Gordon (1971). Cukierman (1974), Turnovsky (1972). and Lahiri (1981). Finally. the expectational variable. P enters the equation. The central proposition of the expectations hypothesis is that the coefficient u, should be unity. This proposition implies that inflationary expectations are fully reflected in wage determiFor a less-than-unitary coefficient. nation. however. we have a partial adjustment for expected rate of price changes. Equation (2) then represents an aggregate demand function derived as a solution to IS and LIM relationships. with the lagged value of c introduced to make the implied consumption function of the permanent-income hypothesis. This equation originates from McCallum (1973). A sketchy derivation of this equation is found in Appendix A to the present paper. The final equation (3) exhibits a general form of the mark-up model on the assumption that the relative share of profits in the value of output is constant.’ Hines (1961). Dick-Mireaux (1961). and Klein and Ball (1959) have tested variants of

this equation to ohtaln rather reasonable rehulrz. It 15 then readily verified that ail three aquationh art‘ oberidentified and it is therefore inappropriate to apply the method of ordinary Iea>t squares (OLS) to an individual equation of thf system. %‘e apply, inbtracl the method of ZSLS to each equation ot the system. A further description of the data \\ill be found in Appendix B. All statistical work has been done with quarterly observations for the period 19651-19S21V. The first quarter of 1965 u’as chosen as the starting point simply because of the availability of consistent data. All variables are expressed as percentage changes from four quarters before.’ Thus R’ = (Lb’, - W,_,)1’11’~_,. etc. While this measure has the advantage of eliminating seasonality cheaply and of reducing the substantial amount of noise rmis-ti-ris quarter-toquarter changes. it has the unfortunate effect that the true residuals will be autocorrelated. As it turns out. all Durbin-Watson statistics obtained by ordinary ZSLS reveal quite strong positi\? auto-correlation. In an effort to mitigate the problem. all the equations are reestimated bv the method of Cochrane-Orcutt 2SLS for a firstorder serial-correlation correction. WC now need to specify P’ - hou inflationar\ expectations are formed. It will be the aim ;f the next section to introduce and explore various versions of the expectations formation scheme> often employed in the literature.

3.

FORMATION

OF ESPECTATIONS

Given that inflationary expectations are largeI> unobservable, researchers have generally been forced to assume unverified models of espectations formation, in which espectrd future intlation is based on actual inflation or other \,ariableb which are currently available.’ fiere a variet\ of possibilities is open, and we experiment with several standard expectations-venerating schemes which fall broadly into two categories: rational and non-rational espectationh. The rational expectations hypothesis simpl! implies that people. in forming expectations. fully incorporate all the economic information at their disposal. On the other hand. the nonrational expectations schemes are based on the assumption that expectations are espressed as more or less complicated functions of past values of price changes. We now start from the non-rational t’y_YXtations schemes. Along these lines, there is a host of possible behavioral hypotheses. LVe propose. inter nlia. the following familiar hypotheses:

F=P_, P = PI, + k(P_, - f$) P=P_,+B(f_,-f-2) P = (1 - b) f-j + 0 P -2 P = f-, + 0 (f-, O
- P-I). 1.

(4) iii

method rational (1961).

(7) (8)

P = E(f+,,

Hypothesis (1). the simplest one can make. is the static expectations hypothesis presumed in the original cobvveb theorem. Hypothesis (5) is the most familiar adaptive expectations hypothesis. according to vvhich expectations are revised by some fraction of most recent forecast error. If the anticipated rate was, say. 5% but the actual rate 10%. the anticipated rate will be revised upward by some fraction J. of the difference between 10 and 5.’ Hypothesis (6) is the estrapolative hypothesis. which asserts that expected inflation equals the past rate of inflation adjusted by a constant proportion 8 of the recent change in the inflation rate. With t) > 0. the forecaster extrapolates the past trend. expecting it to continue. Hypothesis (7) is the weighted expectations hypothesis. according to vvhich espected inflation is a weighted average of actual price changes lagged one and two periods.’ Hypothesis (S) is the normal expectations hypothesis. which postulates that expected intlation equals the past rate of inflation modified by a constant fraction 9 of the difference between the past rate of inflation and the normal level. P_,. towards which people expect the inflation rate itself to adjust eventually. If, for example the recent rate is lower then the normal rate. an increase in inflation is expected for the next period.’ There are several ways of combining the model (l)-(3) with the expectations formation schemes (J)-(S). and of eliminating the unobservable variable P’. Here we follow Solow (1969) and construct a whole time series of P for a given choice of J.. 0, 6, and 9. For the adaptive scheme, in particular, if E, is given, time series of P can be constructed by iteration, starting with a reasonable initial value and using actual series of f.” We use the time series thus constructed for k. 0. o and 9 = 0.1. 0.2, up to 0.9. Some other studies have employed a looser version of the expectations hypothesis by using fitted values of f from the regression. say, f = p,, + p,f_, + ~2 (f-, - P-?) for the extrapolative scheme.’ For the adaptive scheme, in particular, another method commonly used while keeping the original version of the hypothesis, is to apply a Koyck transformation.“’ In applying the rational expectations hypothesis to the model. we follow the estimation

suggested by McCallum expectations hypothesis. specifies that @) = f,,

(lY76).” due to

- ‘1,

The Muth

(9)

where n is a random error (with the classical properties) that is uncorrelated with information @ available and utilized by market participants in forming expectations regarding f,,. The relevant available information, according to the rational expectations hypothesis. consists of present and past values of all the variables in the model (l)-(3) for the present study. In estimating the model according to the rational expectaGons hypothesis, many and varied estimates are possible, the differences reflecting alternative sets of variables in auxiliary regressions used to generate the instrumental variable for P. The drfferent specifications are referred to below Options 1 through 16. The explanatory variables included in auxiliary regressions are indicated for each option in Table I. As the most basic option, Option 1 uses all the predetermined variables in the system (l)-(3). In Options 2-7. alternative combinations of the predetermined variables inclusive of lagged endogenous variables are experimented to see whether estimates of equation (I) are sensitive to the addition of variables. ” Options S-l 1 differ from the above options in that current values of exogenous variables are excluded from the auxiliary regressors. presuming that market participants may not possess information on such values when forming expectations. In Option 12. as an extreme case, actual f,, is used as its own instrument and thus equation (I) is estimated by OLS. Finally, Options 13 through 16 explore the possibility that market participants actually utilize only a portion of the relevant information @ in forming their expectations concerning f,, i.e., market participants are only **partly rational” in forming expectations. This term. introduced by Sargent (1973) is used to refer to a situation in which condition (9) holds but with only the past values of f appearing in @.

4. ESTIMATION

RESULTS

Estimates of the wage equations obtained using Cochrane-Orcutt 2SLS method for alternative models, depending on various hypotheses of expectations formation, are reported in Tables 2 and 3.13 Table 2 shows the results for the nonrational expectations schemes. Mainly in the interest of space, we report only those cases in which h, 8. 6, and 9 = 0.2, 0.4. 0.6. and 0.8.

x x

x x

x x

X

x

x

x

x

x

x

X

x x x

x x x

x x x x x x

x x

X

X

Y

X

X

x

X

X

x x x x x

X

X

X

X

X

x

P-2

x

X

X

x

x

X

X

x

x

X

X

x x x

x x x

X

x

X

X

X

X

x

x

X

X

x

x

X

X

x

x x

X

X

X

X

X

X

X

X

x x

x x

X

x

X

X

X

X

X

X

X

X

X

X

X

X

X

x

X

X

X

X

X

x x x

I’_:

p-1

x

X

x

x

x x

x

x

x

x

x x

x x x

x x

X

X

X

x

X

X

X

X

X

X

x

p-7 p-4

x

x X

*OLS estimates. In virtually all cases, R’ values are more or less reasonable: the Durbin-Watson statistics corrected by Cochrane-Orcutt method are also appropriate. For the variables y and P. each of the coefficients is of the hypothesized sign. and each of the f ratios is reasonably high. The unemployment variable. however, turns out to be insignificant in all the equations estimated. Although we experimented with many possible variants of the unemployment variable. the results remain the same. Th:: result still remains unaltered even when we used some additional labor market variables and treated U-’ as variable following Gordon an endogenous (1976). The statistical insignificance of the unemployment variable in the wage equation. however, is not uncommon or bizarre. Sims (1980). Holms and Smyth (1979) and Lahiri (1981) also report a similar finding.” Most, though not all. coefficients of the expectations variables turn out to be significant; the average of the significant coefficients is 0.4443, ranging from 0.1792 (extrapolative; 0 = 0.6) to 0.7432 (normal; 9 = 0.8). With an average response, these estimates are consistent with those obtained by other studies. Solow (1969). Gordon (1970). and Turnovsky and Wachter (1972) also report similar estimates: they are about 0.4, 0.45. and 0.35. respectively.

The coefficients are always significantly different from unity, which is the value they should take according to the strong version of expectations hypothesis. Accordingly. all versions of the non-rational expectations models we have estimated lead us to reject the strong version of expectations hypothesis, while the results certainly indicate that expectations play an essential role in money-wage determination. This implies. as noted earlier, that wages do not fully adjust to inflationary expectations. Further, a comparison between alternative models of non-rational expectations deserves to be made. While there may be no sufficiently rigorous or sophisticated statistical criterion to choose one of them, the results appear to favor the normal expectations scheme when we consider R’ and significant coefficients of the three simultaneous equations as a whole. A rather pronounced difference is found in the extrapolative models: coefficients of expectations variables are quite low vis-ti-vis those in other models.” The estimation results for the rational expectations models are reported in Table 3. In 16 options we experimented with. all but one coefficient of the expectations variables turn out to be insignificant. An exception is Option 11 in which the auxilary regression includes P-r and P_? only. For the rational models, therefore. even

INFLATION

Table

2. \Yuye rquurions:

199

IN KOREA Non-rurional rxpecrurions R’

A. Staatic

0.8811 (3.1146)”

0.3347 (1.7869)1

0.9575

(2.1903)t 0.6907 (2.1127)t

(3.0989)’ 0.8752 (3.0522)* 0.8485 (3.0696): 0.8558 (3.1053)’

0.7320 (1.6091) 0.5642 (1.6820)t 0.4754 (1.7429)t 0.4037 (1.7786)t

0.0037 (0.7583) 0.0039 (0.8023) 0.0040 (0.8357) 0.0041 (0.8612)

0.6681 (2.0435)t 0.6594 (2.0240)t 0.6522 (2.0091)t 0.6463 (1.9978)t

0.9111 (3.1043)’ 0.9327 (3.0891)’ 0.9484 (3.0762)’ 0.9599 (3.0661);

0.2674 (1.7554)t 0.2169 (1.7207)t 0.1792 (1.6835)t 0.1511 (1.6472)

-0.849 (-0.8802) -0.0876 (-0.9189) -0.0858 (-0.9394) -0.0786 (-0.9073)

0.0032 (0.6501) 0.0032 (0.6615) 0.0039 (0.8219) 0.0046 (1.0272)

0.6839 (2.1008)t 0.6822 (2.138l)t 0.6664 (2.1647)t 0.6427 (2.1636)t

0.8491 (3.1001)’ 0.8311 (3.0114)’ 0.8584 (2.8935)* 0.9146 (2.8657):

0.4043 (1.7776)t 0.4368 (1.6916)t 0.3572 (1.4015) 0.2120 (0.9595)

-0.1037 (-1.0295) -0.1380 (-1.2664) -0.1800 (-1.4812) -0.2024 (-1.5606)

0.0033 (0.6729) 0.0031 (0.6355) 0.0031 (0.6108) 0.0036 (0.7183)

0.7262 (2.1749)t 0.7960 (2.3128)t 0.8723 (2.4385)* 0.8996 (2.5174)’

0.8867 (3.1041)* 0.9039 (3.0881)* 0.9409 (3.0666); 0.9951 (3.0835)*

0.4363 (1.8975)t 0.5797 w;;)t

0.0035

0.6774 (2.0670):

-0.081 I (-0.8442)

(0.7016)

-0.1917 (-1.5202) -0.1309 (-1.2246) -0.1016 (-1.0186) -0.0876 (-0.9017)

0.0038 (0.8041) 0.0034 (0.6954) 0.0032 (0.6552) 0.0033 (0.6612)

0.8461 (2.5368)’ 0.7675 (;:;:;;)t

C. Extrapolative 0 = 0.2 -0.0772 (-0.8130) 0.4 -0.0740 (-0.7866) 0.6 -0.0714 (-0.7655) 0.8 -0.0694 (-0.7489)

B. Adaptive A = 0.2 0.4 0.6 0.8

D. Weighted 6 = 0.2 0.4 0.6 0.8 E. Normal Q = 0.2 0.4 0.6 0.8

(1.9903)‘F 0.7432 (1.7992)t

D.W.

0.5886

1.7922

0.5368

1.9055

0.5792

1.8525

0.5960

1.8110

0.5968

1.7952

O.S762

1.8049

0.5677

1.813-l

0.5616

1.8202

0.5572

1.8257

0.6004

1.7907

0.6046

1.8202

0.5865

1.9059

0.5593

1.9672

0.5821

1.7965

0.5700

1.8079

0.5500

1.8360

0.5242

1.8889

*Significant at a G 0.01. tSignificant at a C 0.05.

the weak version of the expectations hypothesis is rejected. Several interesting facts also emerge from the comparison between the rational and the nonrational models: (i) the R’ values of the rational models are higher than their non-rational counterparts; (ii) the coefficients of actual inflation, P for the rational models are much lower than their non-rational peers; and (iii) coefficients of y are all insignificant in the rational models, contrary to those in the non-rational models.‘6

5. CONCLUSIONS The major results of the present study can be summarized as follows: (l)In the rational as well as the non-rational models, the strong version of the expectations hypothesis cannot be supported. This implies that for the Korean economy, wages do not fully adjust to inflationally expectations. (2)The weak version of expectations hypo-

O.OY-11 (0.9177)

O.IHl24 (0.414)

0.0467 (U. 1996)

5.W

0. 1 IS5

(2.61233‘

(0.6532)

U.7UUS

ct. lOS3

(1.0’)31)

o.Uo24 (U.4377)

0 UW 1 (0.1706)

U.lYj3 (I.l235)*

O.O,S21 (0.3572)

0. hYY4

0.0565 (0.5600)

U.UOZI (O.WJl)

0.109 1 (0.4680)

0.5475 (2.7417)’

0.322-l ( 1.5005)

0.7OSY

4

O.ll.iZ (1.17%)

U.UwS (0.4997)

0.0416 (0.1657)

0.507X (2.4654)*

5

U.1131 (1.1467)

0.00’2 (0.1040)

-0.0159 (-0.0635)

O.-MY (2.1139)’

0. 1122 (0.6SUS)

0 7UOS

6

0. I IS’ (I.IS5l)

O.UOlY (U.3-1lJY)

-0.0073 (-U.37jY)

0.461’) (‘2YUY)t

O.IYll ( 1.0672)

U.7WU

7

0.1121 (1.1372)

o.oU10 (0.365’))

-0.06SY (-0.2709)

0.4765 (‘.369O)i

0. IS71 (O.YY83)

0.7074

x

U.OSYI (0.937s)

0 0042 (0.7SS-t)

0.221x (0.7783)

0.7571 (2.336X)+

-0.32s7 (-U.Y306)

0.702-I

9

U.1191 ( 1.2U-13)

0.007-7 (O.-IS65)

0.0466 (O.lS63)

0.51X7 (z.ZSlh)t

-0.0361 (-0.1407)

O.hYSY

10

0. 1111 (1.1539)

U.0020 (0.363s)

-0.150-l (-O.jS34)

0.3lSS ( I .3S76)

lJ.iS31 (1.5166)

0.700

II

0.1092 (1.1166)

O.OU11 (0.3747)

-0.0696 (-0.17S3)

0.397 I ( I .so47):

0.x-12 (1.1514)

0.704Y

1’

0.1074 ( 1.X5-I)

0.0038 (0.7514)

0.1173 (O.Sl90)

0.33U-l (2.7773)*

0. IO.30 (U.Shl4)

U.6Y7-l

13

-0.0677 (0.703’)

0.003 1 (0.6013)

0.1131 0.4156 (O.-1S7.;,) ( I .YS.ix)+

0.3060 (1.5619)

0.7OY.s

I4

0.0253 (0.2574)

O.OU32. (0.6449)

o.zu59 (O.S61S)

o.-l.i29 (1.2109):

0.457-I (1.6sIo):

0.7105

15

0.053Y (0.5363)

u.uo32 (0.6182)

0. 1704 (U.6940)

0.5073 (Z.jISU)’

O.XJl (1.1402)

0.7042

0.06OS (0.6091)

0.003U (0.5S-11)

0.1437 (0.5972)

U.j(JX2 (2.5222)*

0.2367 ( 1.073(J)

0.7036

1

1 3

16

*Significant tSignificant

I)

-U.O027 (-0.0114)

ChYSS

at a < 0.01. at u s 0.05.

thesis is supported for most non-rational expectations models. while we find the contrary for the rational models. This indicates that only non-rational expectations of inflation play an essential role in money-wage determination. (3)The normal expectations model, among the various non-rational models, is the one which best describes the role of inflationary expectations in the economy. Within the framework of a simultaneous equations model and with a quite general treatment of expectations formation unlike most previous

uork. the present study has been able to generate rather rich evidence on a number of crucial issues about the formation and the role of inflationary expectations for the Korean economy. In particular, the evidence presented above strongly indicates that a single approach of inflationary expectations cannot be viewed as an adequate representation of the complex process involved. For the last two decades Korea has experienced relatively high levels of inflation along with high economic growth. From 1965 to 198-t. the consumer price index increased at an average annual rate of 13.5%.” Inflation began to drop

INFLATION

1s

from lYY1, but the price stability still remains one of the most important policy,objectivcs of the Korean economy. The emptrlcal resu!ts of the present paper about the formation and the role of inflationary expectations for the Korean economy may provide some helpful policy implications to keep inflationary expectatations and inflation itself from soaring again.” Nonetheless. this study remains unsatisfactory, and a great deal has yet to be done for the full assessment of the role and behavior of inflation-

sharply

201

KOREA

ary expectations. In particular. to the extent that the various proxy variables for inflationary expectations are rather naive measures of ‘-true” variables. the conclusions drawn from the present study should be regarded as less than definitive. While this study has produced rather strong evidence on may important issues regarding inflationary expectations for the Korean economy, future work will further refine our understanding of the expectations formation.

NOTES study including ): in the wage equa(1972) for the Japanese economy.

which emerges after the transformation might pick up any serially correlated omitted variables.

2. A brief discussion of this model, coupled with alternative models of price determination. is contained in Jackman ef nl. (1981).

11. A rather detailed discussion on the estimation methods of rational expectations models is found in Begg (1982).

An exception is the unemployment variable u 2. which is a five-quarter moving average of unemployment rate. centered on the current quarter. This is merely a smoothing device to iron out very short-run tluctuations in unemployment.

G , P, are added because these variables’ un&ged counterparts have the greatest explanatory power in the auxiliary regression for Option I.

1. For another tion, see Toyoda

1. Recently several studies use directly observed expectations data such as the data of Carlson and Parkin (1975) on retail prices for the United Kingdom or the Livingston index on consumer prices for the United States. These data are, however, often found to be subject to measurement error. For Korea, no such data are available. 5. The adaptive scheme differs from other nonrational schemes in that it contains an element of learning from past errors while others do not. 6. This corresponds to the case 0 < 0 of the hypothesis (6). in which expectations are said to be “regressive.” See Arrow and Nerlove (1958) on this.

12.

In Option

2, in particular.

three

variables

~1:.

13, Now that our focus centers on the wage equation. estimation results for the income and price equations are relegated to Appendices C and D to this paper. respectively. 1-l. An incomplete but plausible explanation is that the unemployment variable has a rather low variation relative to other variables. The coefficient of variation of ~1 variable is the lowest of all the variables in the model and Cochrane-Orcutt method which involves the first difference of variables might further lower the variation of the u variable. A more detailed discussion is found in Lee (1984).

7. Relative to other schemes, this hypothesis has been seldom employed in the literature. See Gandolfo (1971) for a theoretical discussion on its implications, and Figlewski and Wachtel (1981) for its empirical application. In the present study, as a proxy for the normal inflation rate, we use the average rate of inflation over the past eight quarters.

15. A brief discussion on the estimation results of income and price equations, which are contained in Appcdices C and D. is also in order. For the most part. the R’ values are reasonable: most coefficients turn out to be of hy,pothesized sign and significant. A notable exception IS the government spending variable in income equations. Relative to \vags equations. most coefficients in income and price equations do not change appreciably across different models of expectations.

8. In the present study. the iteration starts from 19611. Solow (1969) and Toyoda (1972) disclose that estimation results are not sensitive to the different starting points of iteration.

16. It is also noted that the results of income and price equations for the rational models do not disclose any marked difference from those for the non-rational models.

9.

See. e.g..

Turnovsky

(1970)

and Lahiri

(1981).

10. As noted in Toyoda (1972). however. this method suffers from the following two flaws: (i) estimation results after Koyck transformation often fail to satisfy sign conditions: (ii) the lagged dependent variable

17. The figures for Japan and Tai\van during same period are 6.7 and 7.6%. respectively.

the

18. A detailed discussion on the policy implications related to the empirical results of this paper is contained in Lee (1984).

312

WORLD

DEVELOPMENT

REFERENCES Arrow. K. J.. and kl. Nerlwc. “A note on expectation5 and the stabtlity ofequil~brium,” Econonwrricn. \‘ol. Z-t (19%). pp. 2s3-2Y3. Begy. D. K. H.. The Ruriod E.vpecrurror~s Rn~oluriotr in :Clocrot,co,fontics~tj~s(Baltimore. hlD: The Johns Hopkins University Press. 1YS2). Carlson. J. A.. and J. 1L1. Parkin, “Inflation expcctations.” Ecorromicu. Vol. 42 (1975). pp. 123-138. Cukierman. A.. “A test of the ‘no trade-off in the long run‘ hypothesis.” Economerrica. Vol. 42 (lY7-1). pp. 1069-1081). Dick-Mireaux. L. A.. “The interrelationship between cost and price changes, 1945-1959.” O.rjkortl Economics Papers. Vol. 13 (lY61). pp. 267-292. Figlewski, S.. and P. Wachtel. “The formation of inflationary expectations,” Review of Ecorromics ad Sttrrurics. Vol. 63 (1981). pp. l-10. Gandolfo, G., Marhernrrricrrl ,Ile~lwtts md Modds irr Ecorrornic Dyrmnics (Amsterdam: North-Holland, 1971). Gordon, R. J.. “The recent acceleration of inflation and its lessons for the future.” Brookirlgs Pq)er.s OH Ecorromic Ah+. Vol. I (1970). pp. ,%41. Gordon, R. J.. “Inflation in recession and recovery.” Brookinys Pupers on Ecotlotnic Acrit?ry. Vol. 7(1971). pp. s-47. Gordon. R. J.. “Comment on paper by W. Oi,” in K. Brunner and A. Meltzer (Eds.), The Econornic.s 01 Price urrd W’uge Corrrrols (Amsterdam: NorthHolland. 1976). Hines. A. G.. “Trade unions and wage inflation in the United Kingdom: 189.3-196 I .” ReLpiew of‘ Economic Srudies, Vol. 31 (1964). pp. 221-252. Helms, J. M.. and D. J. Smyth. “Excess demand for labor, unemployment. and theories of the Phillips curve.” Jowrlcrl of Mucroeconornics. Vol. I (1979). Jackman, R.. C. Mu&y. and J. Trevithick. The Economics of fnflntiorz (Oxford: Martin Robertson. 1981). Klein, L. R.. and R. J. Ball, “Some econometrics of the determination of absolute prices and wages,” Economic Jowd. Vol. 69 (1959). pp. 165482. Kuh. E., “A productivity theory of wage levels - An

APPENDIX

A: DERIVATION

In a rather simple setting, IS-LM relationships be described by the following equations:

can

(Al)

i = g(r) v=c+i

(‘Q)

;w”/P = L(_v. r)

(A-1)

(A3)

IZ

where c = real consumption. )’ = real income, i = real investment. r = interest rate, iv/“ = money demanded, P = price level, and M’ = money supplied. First. assume that the money supply is constant. so that bI’ = ,%I.

Ecw

Srdies.

K., Thr Ecorlornerrrcs of’ Injlariormr~ E.rprcr~(Amsterdam: North-Holland, IYSI). Lee, K. S.. Irzflrlriorrtrry E.vpt~cfcuiom ud Ecorror~ic Sddi~c~ion (in Korean) (Seoul: Korea Development Institute. 19X-l). ‘-Friedman‘s missing equation: McCallum. B. T.. Mardrrsrer Sclrool. \‘ol. 4 1 Another approach.” (1973). pp 311-328. McCallum. B T.. “Rational expcctationr and the natural rate hypothesis: Some consistent cztimates.” Economerricrr. Vol. 11 (1976). pp. lTp;.i2. Muth. J. F.. “Rational expectations and the theory of price movcm2nts.” Ecowtwrricn. k’ol. 2Y (IY61). pp. 315-335. Sargent. 7‘. I.. “Rational expectations. the real rate of the natural rate of unemplo)-mcnt.” intcrest . and Brookirrgs Pcipers Ott Ecotrotnic Acrit,iry. \‘ol. 4 (lY73). pp. 429472. Sim5. C. A.. “hlacroeconomics and realit) .” Econorwrrictr. Vol. -8 (1980). pp. I--IS. Solow. R. hl.. Price Expecrurioru trrltf rhe Brhmt.ior of I/W Price Lewl (Manchester: ,Llanchestcr University Press, 1969). Toyoda. T.. “Price expectations and the short-run and long-run Phillips curves in Japan, lY5&1YhS.” Retfie~ of Ecommics md .Smisrics. Vol. 54 ( lY77). pp. 167-274. Turnovsky. S. T.. “Empirical evidence on the formaJwtn~crl of rile ,itr~enctu~ tion of price expectations.” Srcrrisricd .-1ssociorion, Vol. 65 ( 1970). pp. l-1111454. hhpothehis and Turnovsky. S. T.. “The expectations the aggregate wage equation: Some empirical etiVol. .;Y (1972). pp. dence for Canada.” Ecommici~. l-17. Turnovsky, S. T.. and M. L. IVachter. “a4 tebt of the ‘expectations hypothesis’ using directly observed wage and price expectations.” Rer,iew o_t Ero~~omics und Srurisrics. Vol. 51 (1972). pp. 17-i-1. Wallis, K. F.. Topics in Applied Econornerrics (Osford: Basil Black\vell. 1979). Lahiri.

rims

OF INCOhlE Now

from

EQUATION

(A-l)

and (Ah),

uc

have

(‘~7,

L(y, r) = M/P.

c =f(_v, r)

M’ = h (r) itl” = 1M’ .

to the Phillip cur\e.” RL’I ifti of Vol. 34 (1967). pp. 33.;3hu.

alternative

no,wc

Also

from

(AI)-(

)’ =f(.v,

r) + g(r) = Q(!.

It then

folloas

r=

CAY)

p(y).

Substituting

(AY)

L(_v. 9(y)) = !Il/P. We

(AS)

r).

that

thus have

into (A7)

yields (AIC))

INFLATION

y = ‘o(:CfIP). Now we introduce (A3). First.

(All) two modifications

tn (Al)

c =J_v. r. cc,).

and

(Al’)

following the permanent income hypothesis. (1979). for example. Next.

See Wallis

(A13)

APPENDIX

W: u: y: P: M: G:

and cover

KOREA

203

where G = nominal gwernmcnt spendtn~. Bv performing a similar operation uith (Xl?) and (Al>), we now have a yetwaltzed vcrston of (;\I I): .” =

f(.U. G. P. c-,).

Finally.

by linearizing

(AI_)) (Al-l).

we end up Hith

Y = fi,, + (3,,v + B-G + flrP + p,cc,.

y = c + i + GIP.

All data are quarterly 19651 and 1982IV.

IN

the period

from

Monthly earning in the manufacturing sector, Unemployment rate in the non-farm sector. Output of the non-agricultural sector (1975 = 100). GNP deflator (1975 = 100). Currency + demand deposit = Ml. Government consumption expenditure.

(AIS)

’

B: DATA c: X:

Private consumption Output/employment (1973 = 100). P”: Import prices (1975

expenditure (I975 = lot.)). in the manufacturing srctor =

100).

The source of all variables except n and .I’ is the Bank of Korea; the source of the othrr two variables is the Bureau of Statistics. Economic Planning Board.

WORLD

x-l

APPENDIX

Constant

.bI

DEVELOP.LIENT

C: INCO&lE c

II I Non-rational A.

EQUATIONS P

c-,

RZ

D.b.

0.6260

l.YSIY

0.6230

1 .YS24

U.62Y7

1.9s23

0.62SU

1.YS’3.

0.6’60

1.YS’l

0.6274

l.YSlS

U.hZS7

l.YSlh

O.hZYS

I.YS16

0.6306

I.YSlS

0.6252

1.YS7 1

0.6752

l.Y
0.6262

1.9s23

0.6279

1.9s23

0.6284

1.9SN

0.62Y5

1.982’

0.6275

1.9S24

0.6’13

I.YS26

0.6293

1.9S23

expectations

Sraric 0.0735 (1.7717)t

B. Adtrptive i. = 0.2 O.-l 0.6 0,s

-0.U5S-l (1.107s)

-0.53u1 (-2.316O)i

0.6570 (3.1601)*

0.2158 (3.7345)’ 0.2172 (3.7771)* 0.2167 (3.7674)’ U.‘l61 (3.753s)”

0.0563 (1.0311) 0.0537 (1.0113) 0.05js (1.0603) 0.057s (1.1003)

-0.5421 (-2.403-L)’ -0.5205 (-1.2913): -0.5251 (-2.329l)t -0.530x (-2.3576):

0.6607 (3.2676)* 0.657-l (3.2712)’ U.6554 (3257S)* 0.6554 (3.X5)*

0.2162 (3.75SO)’ 0.2165 (3.7649)* 0.3167 (3.7712)” 0.2169 (3.7765)’

0.0578 (1.0033) 0.0573 (1.079s) 0.0567 (1.067X) 0.0567 (l.lJ578)

-U.5260 (-2.3197)i -0.5719 (-1.3007)‘i -O.SlSS (-2.2S76)‘i -0.515s (-‘.77SY)+

0.6.ilY2 (3.2714)’ 0.6606 (3.x77)* 0.661s (3.2’386)* 0.6627 (3.3073)’

0.07-12 (1.7Y-10)‘; 0.0713 (l.S136)+ 0.0741! (l.S’36)? O.Oi36 (l.S22l)t

0.2158 (3.7493)’ 0.2159 (3.72Y)* 0.2163 (3.7621)* 0.2167 (3.7738)*

0.0533 (1.1126) 0.0577 ( 1.1077) 0.0567 ( I .UYUj) 0.0544 (1.0675)

0.0728 ( 1.7546)t 0.0726 (1.755S)t 0.0739 (1.7975); 0.0766 (1.8766)t

0.2166 (3.7646)’ 0.2170 (3.7718)’ 0.2 168 (3.7614)’ 0.2156 (3.7265)’

0.0564 (1.0652) U.OS4S (1.028X) 0.054S (1.0761) 0.0576 (1.0806)

0.0758 ( I .S344)? 0.0718 (1.756S)i0.0733 (1.7717): 0.0738 (1.783Y)?

C. E.rrrcipolnrire 0.0727 H = 0.2 (1.7517)t 0.07lY 0.1 ( 1.737Y)t 0.0714 0.6 (1.72Sl)t 0.07OY U.S (1.7216)t D. W~iglrrrtt 0 = 0.2 0.4 0.6 0.8 E Normcrl Q = 0.2 U.-l 0.6 0.8

u.2171[2] 0.0729 (1.7573)t ‘Significant tSi%nificant

at a < 0.01. at u < 0.05.

(3.7693)*

-0.5333 -2.3743)? -0.5336 -’ _._iYoU)* -lJ.j308 -2.4100)* -0.5259 -2.4056)’ -0.5236 (-2.3USU)t -0.5205 (-2.2945): -0.5175 (-2.3466)? -0.5-!65 (-2.16SI)*

Rat!T:4;npectations -0.5216 (1.0216) (-2.2Y59)t

0.6551 (3.219-L)* 0.653s (3.2542)* 0.6537 (3.25-c)* 0.654S (3.16S7)” 0.6367 (3.263(J)* 0.6562 (3.1620)’ 0.6539 (3.253S)* 0.655Y (3.2372)’ U.655Y (3.2%2)*

INFLATION

APPENDIX Constant

IN KOREA

D: PRICE W

EQUATIONS p”

x

[ l] Non-rational

R2

D.\V.

0.47T6

1.9416

O.-tS96

1.Y-u.i

0.4795

1.9457

0.4760

1.9157

0.4754

1.9113

0.4750

1.9385

0.3739

1 .Y360

0.4725

1.9342

0.4712

1.9328

0.4747

1.9450

0.472 1

1.9476

0.4681

l.Y-IS7

0.4636

1.9-lso

O.JSOl

1.Y105

0.4855

1.Y3Y1

0.4898

1 .Y386

0.4904

1.9390

0.4623

1.9475

expectations

A. Sttttic

B. Adaptive h = 0.2 O.-l 0.6 0.8

0.0398 (1.0103)

0.5455 (3.4077)’

-0.“56 (-2.G+

0.116J (2.12Ol)t

0.0508 ( 1.3496) 0.0423 ( I .6899): 0.0398 (1.0115) 0.0394 (1.0003)

0.5012 (3.2842)* 0.5340 (3.3850)* 0.5445 (3.4140): 0.5463 (3.4156)’

-0.2211 (-2.5409)’ -(),1771 ____ (-2.4776)* -0.2217 (-2.4578)’ -0.2229 (-2.4700)’

0.1177 (2.219J)t 0.1175 (2.1638); 0.1169 (2.1347)t 0.1166 (2.1257)t

0.5471 (3.4066)* 0.5505 (3.4124)* 0.5544 (3.4211)* 0.05582 (3.430-l)”

-0.2285 (-2.5361)’ -0.2308 (-2.5593)* -0.2325 (-2.5712)* -0.2337 (-2.%32)*

0.1155 (2.1003)t 0.1145 (2.0742)t 0.1134 (2.0479): 0.1125 (2.0241)t

0.0359 (0.9856) 0.0369 (0.9280) 0.0340 (0.8469) 0.0312 (0.7671)

0.5482 (3.4225)* 0.5559 (3.4504)* 0.5674 (3.4849)* 0.5796 (3.5155)’

-0.2222 (-2.45Y3)’ -0.2194 (-2.4150)* -0.2180 (-2.3527)t -0.2184 (-2.3706)t

0.1165 (2.1194)t 0.1155 (2.0903)t 0.1135 (2.0333)t 0.1109 (1.96OO)t

0.0433 (1.1155) 0.0478 (1.2517) 0.0516 (1.3717) 0.0522 (1.3880)

0.5317 (3.3699)* 0.5142 (3.3171)’ 0.4994 (3.2656)’ 0.4976 (3.2545);

-0.2273 (-2.5415)* -0.2292 (-2.5868)’ -0.2307 (-2.6223)’ -0.2306 (-2.6218)’

0.1174 (2.1639)t 0.1182 (2.2084)t 0.1180 (2.2268)t 0.1161 (2.1911);

0.0305 (0.7414)

[2] Rational expectations O.llOil 0.5831 -0.2187 (3.4961)’ (-2.3706)t (1.9350)t

C. E.rtmpoiative 0 = 0.2 0.0397 (1.0039) 0.4 0.0391 (0.9849) 0.6 0.0383 (0.9611) 0.8 0.0376 (0.Y37-1) D. Weighted 6 = 0.2 0.4 0.6 0.8 E. Normal e = 0.2 0.4 0.6 0.8

*Significant tsignificant

at a < 0.01. at a < 0.05.