A Quarterly Econometric Model of Japanese Economy, 1959-1974

A Quarterly Econometric Model of Japanese Economy, 1959-1974

by Koukyu I{AWABATA*

SYNOPSIS

The objective of this paper is to reconstruct a small, quarterly, linear macro-econometric model of Japanese economy so as to clarify the causality among macroeconomic variables and combine the optimal control theory. For this purpose we use a econometric method such as OLS, GLS and TSLS rather than .tne-Kalrnanfilter theory. However, on the point of view of using the medium size digital: computer, we must decrease the size of the model since the dimension of state vector becomes considerably large as the number of lag increases. Consequently, much considera- Lions are given on the rough description of the entire economy rather than the detail description like sectoral and inter-regional model.

Many experimentations are tried so as to realize the both requisitions of logical exactness and statistical significance. Furthermore, total test and final test are employed in order to examine how the model can track former achievement of the Japanese economy.

1. INTRODUCTION

In this paper we construct a small, quarterly, linear macro-econometric model of Japanese economy that can be used to clarify the causality among aggregative economic variables and to combine the optimal control theory which will be discussed in the another paper by the author.

For this purpose, we posed the model into line with Keynesian and post-Keynesian theory and used, in principle, the quarterly, seasonally adjusted data concerning about national income statistics from FY 1959 to 1974. The mode! is a simultaneous equation system with 9 structural equations and 2 identities. In the model there are 35 basic macroeconomic variables of which 11 are endo- genous variables, 8 are exogenous variables and the others, 16 are predetermined endogenous va- riables. which are introduced to explain a dynamic economic bahavior.

As the another research for our problem, however, there are several inspiring studies, such as TCER model, HANDAI model, Klein-Shinkai model and so on. But unlike these full-scale resear- ches, from the point of view of control theory, we are able to consider a econometric system as a linear-discrete-controllable system and to control it optimally to a desirable direction. In other words we can transform the econometric model, B·Yr+P's,=(;r, into the state variable form,^XI<l

=

E(j)=E(

t

r~l

"The Department of Computer Science

whereQand R are positive semi-definite and postive definite rnartices and the symbolE( .J, ~and

.7 denotes the expectation, nominal level and transpose of the vector, respectively. By solving the following dynamic optimization problem:

minimize E[j) = rmrurruze E(~. «X,-XIVQ(Xe-Xr)+(UI-I-ul_dTR(ur-I":' ul-d))

u«,HI, '.', US-l Uo, ill, , ..• US'"'1 l~I

subject to the constraints

xo=⁷⁾

where» is a initial value of state vector, we can stabilize the economic system to a certain equilibrium level. Therefore, to combine the control problem, we must identify the system parameter,B·and F'", by well-known econometric method. Howeverwe must decrease the size of the model since the dimension of x becomes considerably large as the number of lag increases and we are forced to use the medium size digital computer. Therefore we are compelled to reconstruct the short-term econometric model in our own way. Consequently, in this research, much considerations are given on the rough description of the entire economy rather than the detail description like sectoral and inter-regional model.

From the point of view of estimation technique, we can use the another. tool, such 'as the Kalman filter which provides the minimum-variance, unbiased estimator. Since this filter updates the estimates in a recursive way, it assures a practical advantages, But in addition to the for- casting purpose we must clarify the causality among some -significant economic variables and lead it into a reasonable economic theory. From this reason, it will be more advisable for us to use the welt-known econometric method rather than the Kalman filter theory.

On the other hand, the model requires a certain amount of compromize between statistical fitness and theoretical relevance. Especially in case of connecting the estimating theory with the control theory, this cornprornize becomes important and difficult. Therefore, in our mode], many experimentations are tried so as to realize the both requistions of logical exactness and statistical significance.

2. STRUCTURE OF THE MODEL

As \\'3S seen in former section, the model which mainly depends upon the Keynesian and

post-Keynesian theory consists of 9 structural equations and 2 income identities.. The model can be devided. by its nature, roughly into three sectors:(I)the GNP sector, (2) the Financial seelor and (3) the Wage-Price sector. These three sectors are closely interrelated with one another and roughly grasp the actual movement of dynamic economic activity. Furthermore fiscal policy and monetary policy are provided for through exogenous government expenditures GI the change in money supply LlM"

and the change in official discount rale of the Bank of Japan LlRDas a effective policy instrument.

The GNP sector has two important subsectors: (1) the consumption subsector and (2) the investment subsector. The investment subsector, furthermore, is disaggregated and separate equations are estimated to explain gross fixed investment by private enterprises IF, private resi- dential construction IH,privale inventory investmentII, and gross fixed investment by government IG. Then GNP can be defined as

GNp:::: C+IF+IH+II+IG+G+EX-IM

where C is a private consumption expenditure andEX and1Mare exports and imports, respectively.

Disposa ble income YD is defined by

YD =GNP-T ⁽²⁾

where T is total taxes. According to the Keynes' absolute income hypothesis the private consu-

(3)

mption expenditure can be explained largely by disposable income as

c,

=:·il

However, by introducing the Koyck's specification of geometric Jag structure a,= a!1{ i=:1,2,··'

with 01i./3~1, we can transform eq. (3) to an elegant form C.^=: aYDI+{3Ct-1

The first and the second term can be interpreted as the terms which express the income effect and the habit formation, respectively. Although a consumption function such as the above type looks fine when taken by itself in a single equation model, but it is likely to causea instability in simu- ltaneous equations modeL For this reason we must reconstruct the model so as to transfer some of the dependence on income into another explanatory variable. The logical candidate will be a wage rate Wand price level P. Then consumption function to be considered will be of the form

C=:C(YD,C,W,p)

As the gross fixed investment by private enterprises, we can, of course, examine well-known investment function such as acceleration principle, Matthews' capital stock adjustment principle, Kaldor's expanded acceleration principle and Jorgenson's investment function. Because of the risk of straightforward application of these theory into our estimating theory, however, we do not use its directly in the model, but use the slight modification of its. Therefore the function will be of the form

IF = IF(,(JYD,IF, RL )

where LIYD is defined as:L1 YDt;= YDt-YDt-1 ,and RL is the interest rate on loans of all banks.

Similarly private residential construction will be a function of disposable income and short-term interest rate, i.e., of the form

IIJ =: IIJ (l'D, RS,IH) ⁽⁶⁾

where RSis the call money rate. The estimation of private inventory investrnent function is one of most difficult works since it extreamly moves within short periods. Consequently in our small-scale quarterly model we can not necessarily estimate it with a satisfactory form. For this reason we hypothetically use the stock adjustment model and the private inventory stockI([will be

JUt = ^alYDI+a~Ct+aJ]{]t-1

Since the increase of private consumption expenditure leads the decrease of private inventory stock, the coefficient02should be negative. But the other coefficients 01 and 03 should be positive. By taking the first-differencing we get the private inventory investment function

The financial sector consists of l?S and RL functions. Since the short-term interest rate will not depend explicitly upon the level of money supply M, but depend upon the change in money supply LIM, and will correlate the other financial variables, namely, LlRD. then the short-term interest rate function will be

RSt

=:

The coefficients fJ2 and/33 should be positive but the coefficientPIshould be negative since ifLlM increases, tbe interest rate must decrease. The long-term interest rate affects indirectly to an actual economy. And it will depend on changes in income variables ll,YD and also on.JRD and the former value of it. Hence, the function will be as

RL = RL(RS, L1 YD, LJRD, RL) (9)

The wage-price sector consists of three. functions: (1) rate of unemployment VI?, (2) deflator of private consumption expenditure P,and (3) wage rate W. Defining the demand for labour L",

the supply for labour L~,employment E and the level of unemployment U as:

L^D =LD(LlX,J(,W, P) L^S= LSU, L S,U) E= E(L^lJ) and

U = LS-E

where X is output, J{ is total capital stock, and / is a time trend" which implies the population growth. We get

v^{= V(I,LS, V,LlX,K, W, P).}

Then by defining VR,as: UR= U/Ls.and assuming the fact that UR will not contain a time trend, since iti~not influenced by long -terrn population growth, and using the L1 YD and UR instead ofdX and U, we finallyget

UR=^{UR(LlYD,K,lV,P, UR)}

But, since K will be highly correlated with the other explanatory variables in the model, the actual function which should be estimated will be of the form

UR=UR(Ll YD, W, P,VR) (10)

p =P(W, ^YD,IF,^P)

As toP we assume thatP= a'ULe,a> 0and ULe=TV· hi GNP,where tLis hours worked. Then we have

p= rIW+r~JL-rJGNP

with 1'1,,"=.rJ>o. If we assume that the average value of tc is approximately constant, the total hours worked is proportional to the level of employment. ByusingIF andYDinstead ofh andGill?

we get

til)

Finally wage rate functicn is normally assumed to be a function of excess demandL^D_Ls,price level, and rate of unemployment. Bypromoting the similar consideration inUR function we obtai n as a final form:

TV "'"W(L1YD, UR, P,TV) ⁽¹²⁾

On the point of view of model building, much considerations are given on the interactions among three sectors and its relations. The interrelationships between major predetermined and endoge- nous variables in each sector are graphically shown in Chart I. In the next section we describe the equations of quarterly econometric model of Japanese economy which was estimated by using the FACOM~KEMPF/Xapplication program by Fujitsu.

Chart 1. Structure of the Quarterly Macro-econometric Model

(L,,) C'_I .:JYDH IF'-I .:JYD, ,{JJF,_, IH,-, O,IM, 1/,-, 1/,-, YD'-l IG. G. EX, T.

GNP

(En)

C. IF, Ill, II. GNP, YD,

I"

,

I I

IV, .:Jl'D,-,YD, I

P. ~'D'~'^{IF, RS,-, !.1YD,-.}

.:J,'D. ^I

I !

>.1 ~ .,J,

(Ex)11"_1 UR,-,1"-1 (E)()RS<-l .1M, RS,-,

1',-. .dRD, RL'-I

W:J~e-Price Financial

(En) (En)

VR, P.II', RS. RL,

3. QUARTERLY MACRO-ECONOMETRIC MODEL 3.1. Structural Equations

Note the following points:

1. Sample period: 1959-1 to _1971~4 (64 samples) with some exceptions.

2. A suffix(t-j) denotes the- time-lag ofiperiods.

3. Asymbol LJ( ),-ⁱdenotes thet L1 ( )t - i = ( ) I - i - ( ) t-(i"'I).

4. HZ: Coefficient of determination adjusted bydegree of freedom.

S :Standard deviation.

D. W :Durbin-Watson statistic.

The first set of parentheses beneath each estimated coefficient is the I-statistic, and the second is the standard error estimates of the coefficients.

5. In estimating the structural parameterswe used OLS.

1 GNP sector

I-a Consumption subsector Private consumption expenditure

C,=-48907.4 +0.137292 YD,+O.443036Cr-1+806.576W,+ 1000.33P¹ (13) (6,9063) (7.1721) (11.3368)

(7190.52) (0.0199) (0.0618) (71.1465) ]?"l=0.9998 5 =2616.85 D.W. =^1.9460

I-b Investment subsector

(6.3532) (157.4530)

Private fixed investment

IF,:=460.528+a.180586.dYDt-.-t-O .9372691FI-1+0.323981LIYD,+0 .430313L1IF,-1 ([4)

(2.8584) (76.2462) (7.5075) (3.8930) (756.773) (0.06318) (0.0123) (0.o,132) (0.1105)

R'2=0.9973 5=3366.43 D.W.=2.5220 I

Private residential construction

tn,= 1287.69+0.0247438YDI-425.419RS/-J+0.798320IHI-1 (15)

(3.5124) (-1.9790) (12.1404) (2154.14) (0.0070) (214.972) (0.0658)

R²= 0.9937 5 = 2377.19 D.W.= 2.0371 Private inventory investment

n,=-263.168+ 127.8690/+,0.165555IM,+O.390672/{t-l-0.5012i\6'JJ,-~ ^(\5) (2.7609) (3.9132) (3.2589) (-4.3290)

(J984.13) (46.3143) (0.0423) (0.1199) (0.1158)

R^Z= 0.7719 5=6421.77 D. IV.:=1.7677 2 Financial sector

Callmoney rate

RSt=0.661475-0.00003633L1MI+O .921548RSI-l-I-1.4745L:::1RD, (-1.5852) (27.2888) (8.2058) (0.2695) (0.0000) ( 0.0338) (0.1797)

R² =0.9417 5:=0.4495 D.W.= 1.4546

(17)

Interest rate on bank loans

RLt :::: 0.287787+0.0'195988RSt+0 .907329RLr-1+0 .Q0000389.J1"DI - 3+O.229933.JRDt

(1.5378) (22.6164) (3.3941) (5.3856)

(0.25'J4) (0.0109) ( 0.0401) (0.0000) (0.0427)

]?'Z =^0.9716, 5::::0.0823 D.W= 1.7710 3 Wage-Price sector

Rate of unemployment

URr :::: 0.111729+0 .00139,10IV1- 1 - 0.OOOOOZ99.:::11"Dt-3+0 .842270URt-1

(·1.2506) (-2.4299) 06.2465)

(0.05289) (0.0003) (0.0000) (0.0518)

H^Z=0.8706 5 = 0.0812 D.W.== 2.0914

OS)

{I9)

Price

r, :::: 0.8329,12+ 0 . 136763W1- l+'0.000003341"DI - 1+0.959763Pt-1-0.00003429YD/+0 .00007139/Fr (20)

(5.7838) (0.2074) (26.9878) (-2.4386) (3.9324)

(1.6736) (0.0236) (0.0000) ( 0.0356) (0.0000) (0.0000)

HZ ::::0.9984 5::::1.0551 D.W.::::1.5·177 Wage rate

Wr ::::6.46447+^{O.805767W1- I-I.}01717PI-~+ O.895522Pt -1+O.00007653L1 Y Dt+^0.00004330YDr-I (2J)

(15.5109) (-6:8830) (6.286'1) (3.1164) (4.7823)

(3.0949) (0.0519) (0.1478) (0.1425) (0,0000) (0.0000)

!?'Z = 0.9987 5= 1.8288 D.W. = 2.1068 3.2. Identities

1. Gross national product

GNP/= Cr+IFr+IHr+IIrI-IGt+ Gt+ EXt-lMr 2. Disposable income

YDI == GNPt-r,

3.3. Sources of Data

C Private consumption expenditure, 1970 billion yen, Annual Report of National Income Statistics (N.0.

AC Government expenditures, ]970 billion yen, N_L GNPGross national expenditures, 1970 billion yen, N.1.

IF Gross fixed investment by private enterprises, excluding dwellings, 1970 billion yen, N.I.

AIG Gross fixed investment by government, 1970 billion yen, N, 1.

IH Private residential construction, 1970 billion yen, N.I.

II Private inventory investment. 1970 billion yen, N. !.

'"EX Exports, 1970 billion yen, N. 1.

·IM Imports, 1970 billion yen, N. 1.

-M Industrial fund supply, ]970 billion yen, Economic Statistics Monthly: Bank of Japan.

NL Labour force population, thousand of persons, Survey of Labour Force.

'"0 Index of mining and manufacturing production, 1970= 100,Monthly Statistics of MIT!.

P Deflator of private consumption expenditure, ]970= 100, N.1.

•RD Official discount rate of the Bank of Japan, annual rate in%,Bank of Japan.

RL Interest rate on loans of all banks, annual rate in %,Economic Statistics Monthly: Bank of Japan.

RS Call mony rate, annual rate in%.Economic Statistics Monthly: Bank of Japan . .. T Total taxes, billion yen, N. 1.

U Unemployment, thousand persons. Survey of Labour Force, Prime Minister Office.

UR Rate of unemployment,%,UR::= (U/NL)x!OO

IV Nominal wage rate of manufacturing firms. 1970= 100. Annual Report of Labour Force, Ministry of Labour.

YD Disposable income, billion yen. N.!.

denotes an exogenous variable.

4. EXPLANATIONS OF THE l\IODEL

Throughout this section, all of the models are considered from both sides of economic theory and statistical fitness, which is examined by simulation test, and we will extract some interesting results.

Private consumption expenditure

.The private consumption expenditure was estimated in several different forms, all of these forms resulted in the reasonable statistical fit.' The final form was selected more because it gave stability as a whole than for any other reason. The result is shown in eq. (3) and suggests the fairly correctness of our hypothesis which has considered in section 2. Estimated result shows that the short-term marginal propensity to consume is 0.137. This value.is lower than the usual results, such as Ueno's 0.346 and Uchida-Mori's 0.313. This will be due to sample periods and the definition of W.at which the unit of Hiis not a yen but a ratio. But the value of constant term becomes negative.

Much improvement, therefore, will need so as to fit the traditional economic theory. Supposing that previous term's consumption expenditure is written as a distributed lag variable embodying the influence of former icome, we obtain the consumption function:

c,= a+0.137292YDI+O.060825YDt-l+O.026948YDt-z+O.OlJ939 YDt-J+O.005289YDt-~+ ...

From this equation, we can see the influence on the present consumption expenditure of income earned since about 1 quarter past.

An alternative form of private consumption expenditure which was also estimated by OLS was Ct=5858.07 +0. 7343'l3YDI-1-0.699532L1YDt-I-0.957891;1YDt;-J

(57.3751) (-3.5248) (-4.3,170) (30<12.57) (0.0IZ·8) (0.1985) (0.2204) OL5 ]?2==O.9946 5==1316.02 D.W.==O.624

But this was not adapted because the heavy dependence on YD resulted in an unstability in total and final simulation test.

Private fixed investment

Based on eq.(5) the IF function that was estimated at first,by TSLS, was of the form 1Ft== 1491.31+0.11439L1YDr-J+0.9S4615IFr-I-166 .097.1RL-5+0 .478758LlIF,-1

(1.3250) (68.3508) (-0.3037) (3.1121)

(1068.94) (0.0863) (0.0144) (546.974) (0.1538) TSLS R² =0.9948 5::=4659.89 D.W.==1.8830

The coefficients of L1YDt-~ and iJRLr-5 are not significant. Therefore, by dropping the LlRLt-5, we reestimated the function, and the final function that was adapted after several experimentations was eq.fl-l). As a results. the explanatory power has considerably risen and all of the coefficients have satisfied the logical sign test.

Private residential construction

The result of private residential construction function is good. As was seen in former consi- deration, because of the illness of housing institution of government. disposable income and short- term interest rate have considered as a major explanatory variables and have shown a statistical signiIicance.

Private inventory investment

Because ofextrearn fluctuation in each phase of business cycle the estimation of this function is one of the most difficult works in macro-econometric model building. The primary approach was I It= - 52.3682+0.031139YD^{1 -}0.118133LJ'Ct-~+O.432186!II-! - O.OZ23863L1YDI-~ (22)

(2.8136) (-0.6938) (2.759 ) (-0.2838)

0964.22) (0.0110) (0.1703) (0.1566) (0.0789) OLS H'2= 0.6699 S =7718.17 D.W.= 1.6275

As was anticipated, the statistical fitness was considerably worse. The coefficients of LJ'Ct-~ and LJ'_YD,-~ yielded a negative and insignificant coefficients. There was a tendency for YDI and L1 YD^{I -}z to become multicolinearity. Hence, as secondary approach, we dropped LlYDr-z and, to rise Durbin-Watson statistic. we used the GLS instead of OLS. The model becomes

n,= -298.480+ 0.0352376YDt-O.165588L1C'-2+0 .36213HII-J

(3.4156) (-0.9384) (2.3636)

(1950.71) (0.0103) (0.1765) (0.1532) GLS ]?2=0.6022 5 =0.7760 D.W= 1.6907

Like eq. (22) the coefficient of LlC:-~ was not significant. Hence, we abandoned the use of£1Ct-2 and adapted the new variable 0, which was highly correlated with Il. Throughout the several experimentation the final form was obtained in eq.tlb),

Call money rate

The short-term interest rate was explained, as was considered in former section by LJ'M, LJRD • and RSr-1and the result gave better statistical fit.

Interest rate on bank loans

Interest rate on bank loans function was a function of call money rate, changes in official discount rate of the bank of Japan, and change in disposable income lagged three quarters.

Rate of unemployment

Though the hypothesis suggested that the price level should be used as a important explanatory variable, the result yielded the unfittable sign and statistical insignificance with respect to it. Hence, we dropped it and after several experimentations we get the eq.(19). The other form including the price level is shown as a reference:

URI= -0.00337768+0.0000000029£1YDI-,+a.0000164734lVt-1 +0;856744URI-l

(0.2682) (1.3462) (15.8242)

(0.0015) (0.0000) (0.0000) (0,0541)

+0.00008507Pt-J-O .OOOOOOO,7104IGr (3.<1110) (-3.3744) (0.0000) (0.0000)

GLS HZ=0.8921 5 =0.9145 D.W.= 2.1778 Price

The price function contained a dependence on the disposable, income and private fixed invest- ment and wage rate lagged one quarter. It seemed that' there was a tendency for YDrandYD^{t - 1} to become multicolinearity. Hence we dropped the YDr-1 and reestimated the function. But the result does not yield a reasonable fitness. Another form was tried by adding the private inventory invstment lagged two quarters:

r.

(5.5562) (0.3962) (-0.6791)

(1.7485) (0.0262) (0.0000) (0.0000)

-0,0000384099YDt+O.0000729314IFr

(-2.4617) (3.6938)

GO.OOOO) (0.0000)

GLS HZ = 0.9982 5 = 0.9991 D.W.= ].6448

(24.4990) (0.0387)

But the result becomes worse.

Wage rate

It should be noted' that the rate of unemployment was deleted in the model. though it will be a important explanatory variable, and all of the coefficients have a statistical significance. However, from the point of view of economic theory, it had better to employ the rate of unemployment and reestimate the structural equation.

To examine how the simultaneous model can track past achievement of Japanese economy total test and final test have been conducted. FY 1959's datas were given as initial values in the tests.

The results are shown in Table 1 and Chart 2.

Table 1. Results of Simulation Test Variables

C IF IH II

RS RL UR

P W

GNP YD

Total Test 0.9996 0.9975 0.9938 0.7864 0.9445 0.9719 0.8768 0.9987 0.9988 0.9995 0.9991

FinalTest 0.9898 0.99G8 0.9763 0.7130 0.7692 0.6269 0.6273 0.9831 0.9855 0.9952 0.9925

The figures designate the values of coefficient of dctcrm inetion

Undoubtedly, most of the results of simulation test are satisfactory except for the private inventory investment, call money rate, interest rate on bank loans and the rate of unemployment.

Furthermore from the final test we can see the tendency of slight overestimation on price level during 1959-1 to 1962-4. 1964-1 to 1964-3 and 1969-4 to 1974-4.

5. CONCLUSION

Throughout this paper we reconstructed the quarterly linear macro-econometric model of Japanese economy in order to combine the control theory and to use a medium size digital computer.

The sample period was 1959-1 to 1974-4 (64 samples) and it was larger than the other model such as SP-15 of EPA. This is mainly due to obtaining the fittable interpolation test. In other words, as can be seen in the literature of modern control theory, the changes in parameter estimates have a large impact on the optimal state trajectories and especially optimal policy is more sensitive to the changes in a coefficient of autoregressive term than the changes of the others. Consequently to get a stable parameters the sample size became larger.

As to estimation techniques, TSLS was, of course, considered so as to get a consistent estima- tors. But the results by applying the TSLS caused an unstable properties and there was a high poss- ibility that this facts was strengthened in proportion to the size of the model and to the complexity of the specification of the model. Therefore we used the OLS instead of TSLS for equations having unstable parameters. As a results, though, in general, it was said that applying OLS in estimating simultaneous equation system tend to bear a upper bias, the results of final test was satisfactory. From this fact we should examine the efficiency of TSLS estimates in relation to the other estimates such as OLS, GLS and LlML etc..

Chart 2. Graphs of Final Test

c IF

___ Actual

""" Estimale billion yen

-!("L'),u);r---- - ' - - - ,

...:r,«(0)

Fiscal Year

billion yen

:",.c".",---'---__

Fiscal Ye;H

fH billionyClI

r:.~"r_-....:....---

..

!,.~ i.I~ (,-;' M u ;~ ~1 ~ n ^':'t PisG.11 Year

Annual rate % RS

','Co;f)

billion yen

AnnualralC%

II

RL

;-: n ':l.

fisC<11Year

~ ~ ~l ~ u ~ d U ~ ~ b ~ ~ ~ ~ ~4 fiscal Year

~-_.~...-t-, -,

~~ n .3 7c

Fiscal Year

'0i\nnualfale % UR _19iO=IOO ^p

(r) (.~ ~··r .1,. .f;-;'" ..~ ~.~ ::':) 71 n n H

Fiscal Year

f..) ,:..a I!-:' v:' j.': r:... r::t :'":l 11 i~ 1'l i~

Fbc.ll Year

rv _GNP

19;0'" 100

:C~r---, billiony,"IJ

l~.l"''')'----'---_-''

.../ ...,.

...•.

l.) ~. ~ t~ .;.~ !:o:'i '0 ::"J) 'a 7:~:-) :t

FIscal Year ^';1Fiscal Year^{;-: ';") ;,}

billion yen YD

ot:-:' (.{; G'7 ';Ii ":1 rc ~I ;-.: :'"l :--l

FiscalYear

Because of too heavy consideration of using the medium size digital computer, the model was extreamly crushed ;nto small test model rather than practical model. However, 011 the other hand, our model will have the special merit. In other words we will be able to manange the national economy so as to achieve stable economic growth without serious inflation and high rate of unemp- loyment by solving the optimal control problem, which has considered in sec Lion 1. Byexamining the various simulation experimentations we also attain a optimal policy. but this is to some extent indirect and trial and error in nature.

Despite the fact that the actual economy has a dynamic and nonlinear structure, modern econometric method will be a effective approach on clarifing the interdependence of various econo- mic variables and on. Iorcasting the characteristics of national economy. But from the point of view of theoretical side. the econometric method will be more fittable 00c1arifing the structural interde- pendence rather than on Iorcasting the system behavior. On the contrary, the Kalman filter theory will have a inverse tendency. The method satisfying both requisites is still, however, unsettled,

For our fruitful research, it will be necessary for us to expand the scale including the trade sector and the government sector and so on. Furthermore it is also interesting to construct a sectoral and inter-regional model, and to control the model to follow the ideal path.

ACKNOWLEDGEI\IENT

The author is grateful to Kazuhiro Konomi of Kyushu Institute of Technology who did the computer operation with great competence and enthusiasm.

REFERENCES

( 1 ) Allen,RG.D. :Macro-economic Theory. Macmillan and Company Limited, 1967.

(2) Economic Planning Agency; Tanki Kcizai Yosoku Master II/Iodel no Kenkyu. Kcnkyu Series, 21. 1968.

(3) Economic Planning Agency: Economic Models for Basic Economic and Social Plan, 1973-1977, Economic Planning Agency, Government of Japan, 1973,

( 4 ) Economic Planning Agency: Tanki Kcizai Yosoku Pilot Model SP-IS. Kcizai Bunseki, 52, 197'1.

( 5) Evans. M.K :Macroeconomic Activity.Harper & Row, Publishers, 1959.

( G ) Jorgenson, D,W. andj.A.Stephenson :,"Invcstment Behavior in U.S. Manufacturing. 19'17-1960,"Econometrica.

35-2,( 1%7),169-220.

( 7) Kuwabata.K.:" Studies on the Linear Regression Analysis with Error in Variables," Memoirs of the Kyushu Institute of Technology, Engineering, 5(1975), -13-58.

( 8) Kawabata, K. : "Studies on the Lag Structures in the Economic Dynamics and their Statistical Estimation,"

'Bulletin of the Kyushu Institute of Technology, Humanities, Social Science, 24(1976),45~62.

( 9JPindyck, RS. : Optimal Planning for Economic Stabilization : the Application of Control Theory to Stabilization Policy. North-Holland Pub. Co., 1973

[10) Taternoto, M. and S. Ichirnura : Ninon Keizai no Kciryo BunsekiTouyoukeizai, 1970.