Export-Led Growth in Malaysian Agriculture: A VAR Approach

Pertanika J. Soc. Sci. & Hum. 1(1): 63-69 (1993) ISSN: 0128-7702 © Universiti Pertanian Malaysia Press Export-Led Growth in Malaysian Agriculture: A...
Author: Basil Hall
2 downloads 0 Views 329KB Size
Pertanika J. Soc. Sci. & Hum. 1(1): 63-69 (1993)

ISSN: 0128-7702 © Universiti Pertanian Malaysia Press

Export-Led Growth in Malaysian Agriculture: A VAR Approach AHMAD ZAINUDDIN ABDULLAH Faculty o f E conom ics and M anagem ent, Universiti Pertanian Malaysia, 43400 UPM Serdang, Selangor Darul Ehsan, Malaysia.

Keywords: Causality, growth, variance decomposition, VAR ABSTRAK Kertas ini mengkaji hubungan diantara eksport-pertumbuhan bagi sektor pertanian dengan menggunakan m odel tiga-angkubah vektor autoregressi (VAR). Model ini telah diuji dengan tiga prosedur penyebab yang berbeza : penyebab Granger, teknik Hsiao dan varians dekomposisi. Hasil kajian menunjukkan pertumbuhan keluaran dalam negeri kasar menyebabkan eksport dalam dua daripada tiga ujian yang digunakan.

ABSTRACT This paper investigates the export-growth relationship for the Malaysian agricultural sector using a threevariable vector autoregressive (VAR) m odel. The m odel was subjected to three different causality tests procedure: Granger causality, H siao’s technique and variance decomposition. The results indicate that growth o f gross domestic product (GDP) causes exports in two o f the three test procedures employed.

INTRODUCTION T h e causal re la tio n s h ip b etw een a g ric u ltu ra l exports and growth in Malaysia is investigated in this paper. O ver the years, th e co n trib u tio n of agriculture has been declining and was eventually surpassed by the m anufacturing sector in 1987. Despite the decline, agriculture is likely to rem ain im p o rta n t. T h e fo rm u la tio n o f th e n a tio n a l agriculture policy, an d its su b seq u en t revision, clearly underscores the im portance of this sector. T h e p r e s e n t a p p ro a c h e m p h a s is e s m o re o n capital-intensive operations which would at least m a in ta in a g r i c u l t u r e ’s c o n tr ib u tio n to th e Malaysian GDP. It is known th at agriculture in Malaysia is a labour intensive industry. In itia lly M alaysia a llo w e d an im p o rtsubstitution strategy to realize a high growth rate. Subsequendy it followed the recom m endations of the OECD and NBER studies o f the 1960’s and 1970’s w hich suggested th a t an im p o rt substi­ tution policy would show significantly less growth than would a policy based on export prom otion.1 T he country then em barked on a strategy based on export prom otion with confidence.

P rev io u s stu d ie s on th e e x p o rt-g ro w th relationship were based on cross-country studies, and on time series studies which produced mixed re s u lts. In th e case o f M alaysia, a stu d y by Habibullah and Yusoff (1990) concluded that the agricultural sector causes growth. However, this result and o th e r studies based on the causality m ethod have to be in te rp re ted with great care before any general causal ordering can be made. Most of the previous studies based their results on contem poraneous correlations and regressions. A high correlation between two variables does not n e c e ssa rily im ply th e e x is te n c e o f a c au sal relationship between them. In fact, an observed high correlation between exports and growth is n o t surprising, as exports are a co m p o n e n t o f a g g re g a te d e m a n d in a K ey n esian id e n tity . E x p o rts c o n trib u te to e co n o m ic grow th only th ro u g h cap ital fo rm a tio n (K avoussi, 1984). Therefore, the direction of causation is difficult to ju d g e. T he establishm ent of causal o rd erin g is very im p o rta n t fo r less d e v e lo p e d c o u n trie s (LDCs). If export (X) expansion is found to cause growth (Y), then export led growth is favourable.

1 See Little, Scitovsky and Scott (1970) for OECD study. Bhagwati (1978) and Krueger (1978) are synthesis o f the NBER study.

Ahmad Zainuddin Abdullah

O n the o th er hand, if reverse causality is found, the less developed co u n tries have to achieve a threshold developm ent before ex p anding their e x p o rt secto r (M ichaely, 1977). B id irectio n al c a u sa lity in d ic a te s th a t e x p o rts a n d g ro w th reinforce each other. Results based on a specific country study are very im portant as the adoption of a wrong policy can prove cosdy to a developing country. In fact, Sharm a et al (1991) found that there were no two identical results am ong the five industrialized countries they studied. T o d ate a n u m b e r o f studies have investi­ g ated th e d ire c tio n o f causality in a bivariate c o n tex t (Jung an d M arshall 1985, Chow 1987, Kwan and Cotsomius 1990, H abibullah and Yusoff 1990), with very few using muluvariate framework (Kunst and Marin, 1989, Sharm a et a l 1991). This study em ploys a v ecto r au to reg ressiv e (VAR) technique to investigate the nature of the causal relationship between exports (X) and growth (Y). To remove spurious correlation, the direction of causation betw een ex p o rt an d grow th is invest­ igated with the presence o f a third variable Z. The p aper is organized as follows: Section II d e sc rib e s th e causal re la tio n s h ip am o n g th e v a ria b le s in c lu d e d in th e study; S e c tio n III presents the m ethodology. Section IV describes th e d a ta a n d d isc u sse s th e re s u lts , w h ich is followed by a b rief sum m ary and conclusion in Section V.

CAUSALITY: THEORETICAL ANALYSIS T h e r e a re tw o a p p r o a c h e s to e m p ir ic a l investigation o f the relationship between export e x p a n sio n a n d eco n o m ic grow th. T h e first is based on Ram's (1985) claim th at exports can have a positive im pact on econom ic growth due to a b e tte r a llo c a tio n o f re s o u rc e s a n d th a t exports also can cause econom ies o f scale an d externalities an d stim ulate grow th2. T h e o th e r a p p r o a c h is b a se d o n a “tw o -g a p ” m o d e l o f g ro w th w h e re in c re a s e in e x p o rts cau ses an increase in im ported capital goods which in turn raises the growth rate o f capital form ation and th u s s tim u la te s g ro w th (V o iv o d as 1973;

W illiam son 1978; Fajana 1979). W hile e x p o rt expansion can lead to growth, it is also plausible that econom ic growth causes export expansion*. Recently, H elpm an an d K rugm an (1985) have su g g e ste d a b id ir e c tio n a l c a u sality b e tw e e n e x p o rt a n d grow th. A ccording to this theory, ra p id g ro w th lead s to e ffic ie n t a llo c a tio n o f reso u rce s d u e to c o m p arativ e ad v an tag e a n d allows for the exploitation of econom ies of scale. O nce econom ies of scale are realized, the costs of exportable goods will decline and hence exports will be m ore com petitive in the w orld m arket. T h e re fo re th e causal re la tio n sh ip may ru n in b o th d ir e c tio n s a n d o fte n te n d s to be se lf reinforcing4. This suggests that factors other than exports can also cause growth. In this study, capital is in cluded as a third variable that could explain growth. Traditionally, on the basis of neoclassical growth theories, it is believed that capital stocks lead to o u tput growth, which in turn leads to further capital form ation via th e a c c e le ra tio n p ro c e ss. In th e case o f Malaysia, we may consider capital as endogenous to the grow th process an d the lab o u r in p u t as exogenous as there is a surplus of labour in the economy.

METHODOLOGY AND VAR SPECIFICATION A wide range of studies in economics have used Granger-causality (1969) tests". The central them e of this test is that a variable X is said to cause Y if Y is b e tte r p re d ic te d by u sin g past values o f X (which are contained in the inform ation set that includes both X and Y) than by not using them. In order to draw a m eaningful causal link between X and Y, one m ust consider as m any factors as p o ssib le in th e in f o r m a tio n set. T h e s e m ay include such internal factors as composition and direc tio n o f exports, investm ent activities an d external factors such as the econom ic growth of d e v e lo p e d c o u n trie s a n d so o n. M ost o f th e studies m entioned above use G ranger tests in a b iv a ria te c o n te x t. E x te n d in g th e te s t to a multivariate framework involves the inclusion of the new variables and lag associated with them ,

2 Views shared by Tyler (1981); Feder (1983). 3 See Jung and Marshall (1985). 1 Kunst and Marin, (1989), and Jung and Marshall, (1985) did not support this hypothesis. See Sims (1972) money and income, public expenditures and national income, Singh and Sahni (1984), and Islam and Rafiquzzaman (1991) who investigate property tax and intermunicipal migration relationship in Canada. 64

PertanikaJ. Soc. Sci. & Hum. Vol. 1 No. 1 1993

Export-Led Growth in Malaysian Agriculture: A Var Approach

which exhausts th e d egrees o f freed o m ra th e r q u ick ly . B u t as n o te d by H sia o (1 9 8 2 ) a n d L u th k e p o h l (1 9 8 2 ), th e ex clu sio n o f a th ird variable may lead to spurious correlations. H ence in this paper, a multivariate framework is used to in v e stig a te th e c au sal r e la tio n s h ip b e tw e e n e x p o rts a n d grow th in M alaysia. As n o te d by S h a rm a et al. (1 9 9 1 ), a VAR m e th o d has an a d v a n ta g e o v er o th e r te c h n iq u e s b e c a u se it co nsiders all possible causal influ en ces o f the variables included in the system. A sp e c ific a tio n o f a th r e e v a ria b le VAR m odel is expressed as:

o f n o n -statio n ary series may lead to sp u rio u s inferences. In this study, a standard tool in time series analysis is used to convert the data into sta tio n a ry tim e series. H ow ever, th e issue o f determ ining the optimal lag length in Model (1) is still elusive'1. T he sequential m ethod suggested by Hsiao (1979, 1981) which com bines Granger causality and Akaike’s m inim um final prediction error(FPE) is used to determ ine the optimal lag. H siao (1981) has n o te d th a t use o f th e FPE b alan ces th e risk o f bias w hen a low er lag is ch osen against th e risk o f in crease d variance when a higher order is chosen. Furtherm ore, this technique does not constrain the lag from being the same and it is also equivalent to applying an F -test w ith v ary in g s ig n ific a n c e levels. T h e procedure to determ ine the optim al lag length for each variable is outlined below :

The entry cp'ij(L) has the following interpretation. T h e su p e rsc rip t I in d icates th e optim al lag o f variable j in equation i and L is a lag operator. (pj() (i = 1, 2, 3) and Jlit (i = 1, 2, 3) are constant and white noise e rro r term s respectively. To test for prima-facie causation between ilh and j h variables, zero restrictions on param eters are tested. For example, j h variable prima-facie cause i,h variable if and only if (p.. * 0, and the i,h variable is said to prima-facie cause j ,h variable if (p. * 0. Hsiao (1982) has noted that if the j lh variable prima-facie causes the k,h variable and klh variable prima-facie causes th e ilh variable, th e n th e j ,b variable prima-facie causes the i,h variable indirectly. Thus, the model a c c o u n ts fo r b o th d ire c t a n d in d ire c t causal relationship in the variable of interest. The model is derived from a covariance stationary process which has a constant m ean and autocovariances. O ne of the main requirem ents in applying a G ranger-causality test is th e stationarity o f the data. For the m odel to be covariance stationary, the tim e series considered m ust be constant in b o th m ean a n d a u to c o v a ria n c e s. T o achieve stationarity, the data are filtered by using suitable m ethods (e.g. Sims’ filter, H siao’s technique, first d ifferen cin g , etc.). O therw ise, draw ing causal in f lu e n c e s will p r e s e n t p o te n tia l p ro b le m s (G ranger and Newbold 1974). A lthough several VAR s tu d ie s have u s e d n o n -s ta tio n a ry d a ta directly, O hanian (1988) has shown that the use

i)

X, = FPE (q, r) th en X prima-facie causes Y and thus variable X is added to the equation (2). S im ilar step s a re ta k e n fo r v ariab le Z in equation 2. Sharm a et al. (1991) have noted that VAR analysis is sensitive to the ord er o f the variables. H ence the above procedure is a p p ro p ria te fo r re d u c in g any bias arising from m isordering o f the variables. A fter th e analysis from step (i) to (iv) is perform ed, all the equations are estim ated using Z e lln e r’s (1962) iterative seem ingly unrelated regressions technique.

M ost o f th e d a ta analyses u sin g th e VAR technique are centered on variance decom posi­ tion and im pulse response functions which are g e n e ra te d from th e m oving average re p re se n ­ tation of an autoregressive process. The movingaverage representation of a VAR model involves a lin e a r c o m b in a tio n o f past a n d c u rre n t in n o ­ vations o f th e v ariab les in th e system . If th e innovations are co ntem poraneously correlated, th e m o d e l’s v a ria n c e can be d e c o m p o se d . A standard way of doing this is to orthogonalize the jnit in (1 ). Sim s (1980, 1982) has n o te d th a t variance decom position is useful in checking the causal influences in the system, and in fact the strength o f G ranger-causality can be m easured through this decom position. It decom poses the variance due to the innovations in own variables as well as o th e r variables. A variable is strictly

exogenous if it is 100% due to its own innovation. On the other hand, if its variation is partly due to innovations of another variable, there is evidence o f weak c a u satio n . For e x am p le, if grow th is ex p lain e d by only a small p o rtio n o f a n o th e r variable’s forecast erro r variance, then it is a case of weak prima-facie cause. If the result of a VAR variance decom position is sensitive to the o rder o f th e variables, th e specific gravity c rite rio n described in step (iv) is used to guide the order­ ing o f th e v ariab le s. H e n c e th e r e a re th r e e d if f e r e n t o r d e r in g s fo r th e v a ria n c e decomposition.

DATA REQUIREMENTS AND ESTIMATION RESULTS A nnual data for GDP, a g ricu ltu re e x p o rt an d gross capital form ation for the 1960-1989 period u s e d in th is stu d y a re ta k e n fro m th e I n te r n a tio n a l F in a n c ia l S ta tistic s Y ea rb o o k (1990) and W orld Tables (various issues). T he shares o f export and gross capital form ation are th e n u s e d to in v e s tig a te th e r e la tio n s h ip b e tw e e n th e v a ria b le s . As c a u sa lity te s tin g r e q u ir e s s ta tio n a r y d a ta , th e B o x -Je n k in s technique was used. To d eterm ine the optim al lag for each series, we then em ployed Akaike’s FPE. For exam ple, for agricultural exports, we found the optim al lags for each series are 1 for Y, 1 fo r X a n d 1 fo r Z. F o llo w in g H s ia o ’s technique, these are then treated as controlled variables while variables X and Z, Y and Z and X and Y are m anipulated variables in equations Y, X and Z respectively. Based on this optim al lag we in v e stig a te d G r a n g e r ’s c au sa lity in b o th b iv a r ia te a n d tr iv a r ia te c o n te x ts , H s ia o ’s causality and Sim ’s variance decom position. As th e fo recast e rro r variance is sensitive to the o rd e r of the variables, the above p ro c e d u re is used to o rd e r th e variables in each e q u atio n . T h e fin a l s p e c if ie d m o d e l fo r a g r ic u ltu r a l primary exports is

X, X

* Significant at 10% level

Hsiao’s Sequential Technique FPE Comparison

Conclusion

FPExv(q, r) < FPE (q)

Y causes X

FPEx/(q, r) < FPEx(q)

Z causes X

F P E Jq, r) < FPEv(q)

Y does not cause Y

FPEv/(q, r) < FPEv(q)

Z does not cause Y

FPE x(q, r) < FPE (q)

X does not cause Z

FPEx(q, r) < FPE (q)

Y does not cause Z

TABLE 2 FEV decomposition for agricultural export Variables Steps Explained Ahead

Explained by innovations in Y X Z

Y

8

100.00

0.00

0.00

X

8

14.37

81.30

4.33

z

8

0.00

0.00

100.00

ACKNOWLEDGEMENTS The author is indebted to Dr. N. Islam and Dr. Douglas Willson of C oncordia University and Dr.

Pertanika J. Soc. Sci. & Hum. Vol. 1 No. 1 1993

67

Ahmad Zainuddin Abdullah

R o slan A. G h a ffa r o f U n iv e rs iti P e r ta n ia n Malaysia for constructive com m ents on earlier versions o f this paper.

W.S. and P.J. M a r s h a l l. 1985. Exports, growth and causality in developing countries. J. Dev. Econs. 18: 1-12.

Jung,

R.M. 1984. E x p o rt e x p a n sio n and economic growth: further empirical evidence./. Dev. Econs. 14: 241-250.

K a v o u s s i,

REFERENCES J.N. 1978. Foreign trade and economic development: anatomy and consequences of exchange control regimes. New York: NBER.

B h a g w a ti,

C a in e s ,

P.E.,

C a u s a lity

C .W . K e n g , a n d S . P. S e t h i . a n a ly s is

and

1981.

m u ltiv a r ia te

a u to r e g r e s siv e m o d e llin g w ith a n a p p lic a tio n to s u p e r m a r k e t sa le s a n a ly sis.

A. 1978. Foreign T rad e regim es and economic development: liberalization attempts and consequences. Cambridge. Mass: Ballinger NBER.

K rueger,

and D. M a r in . 1989. On exports and productivity: A causal analysia. Rev. Econs. Stats. 71: 699-703.

K u n s t , R .M .

Econ. Dynamics and

Control. 3: 267-298.

C.Y. 1987. Causality between export, growth and industrial development: empirical evidence from the NICs./. Dev. Econs. 26: 55-63.

C h ow , P e te r

O. 1979. Trade and growth: The Nigerian experience. World Development. 7(1): 73-78.

A.C.C. and J.A. C o t s o m it is . 1990. Economic growth and expanding export sector: China 1952-1985. International Economic J. 5(1): 105116.

K w an,

F a ja n a ,

F eder,

I., T. S c ito v s k y , and M. S c o t t . 1970. Industry and Trade in Some D eveloping Countries: A Comparative Study. London: Oxford University

L it t le ,

G. 1983. On Exports and economic grow th./

Dev. Econs. 12: 59-73.

Press.

C.W.J. 1969. Investigating causal relations by eco n o m etric m odels and gross spectral methods. Econometrica. 37: 424-438.

G ranger,

C.W.J. and P. N e w b o ld . 1974. Spurious regressions in econometrics. J. Econometrics. 2:

H. 1982. Non-causality due to ommitted variables. / Econometrics. 19: 367-378.

L u t k e p o h l,

1977. E xports an d grow th: An empirical investigation./ Dev. Econs. 4: 49-53.

M ic h a e ly , M.

G ranger,

L.E. 1988. The spurious effects of unit roots on vector autoregressions: A Monte Carlo study. / Econometrics. 39: 393*414.

111- 120 .

O h a n ia n ,

and Y u s o f f , M oh am m ad B. 1990. Exports and economic growth. A Case o f M alaysia. In Trade and Development in Malaysia, ed Mohammad Yusof et al. Serdang: Universiti Pertanian Malaysia Press, p. 83-92.

Ram, R.

H a b ib u lia h , M u z a f a r S h a h

E. an d P.R. K r u g m a n . 1985. Market Structure and Foreign Trade. Cambridge: Mass.

H e i.p m a n ,

1985. Exports and economic growth : some additional evidence. Economic Development and Cultural Change. 33: 393-414.

S.C., M. N o r r i s and D.W. C h e u n g . 1991. Export and economic growth in industrialized countries. Applied Econs. 23: 697-708.

S h arm a,

MIT Press. C. 1979. A u to reg ressiv e m o d e llin g o f Canadian m oney and incom e data. J. Amer. Statistical Assn., 74: 553-560.

H s ia o ,

Econ. Rev. 62: 540-552. S im s ,

S im s,

C. 1982. Autoregressive modelling and causal o rd e rin g o f e c o n o m ic variab les. / Econs Dynamics and Control 4: 243-249.

S in g h ,

H s ia o ,

M.N. and M. R a fiq u z z a m a n . 1991. Property tax and inter-municipal migration in Canada : A multivariate test of the Tiebout Hyphothesis. Applied Economics, 23: 623-630.

Isia m ,

C.A. 1980. M acroeconom ics and reality.

Econometrica. 48: 1-48.

C. 1981. A utoregressive m odelling and money income causality detection. / Monetary Economics. 7: 85-106.

H s ia o ,

68

C.A. 1972. Money, income, and causality. Amer.

Sims,

C.A. 1982. Policy analysis with econom etric models. Brookings Papers on Ecnomic Activity. 1: 107-152.

B. B.S. S a h n i. 1984. Causality between public expenditure and national income. Rev. Econs. Stats. 66: 630-644.

W.G. 1981. Growth and export expansion in developing countries: some empirical evidence. / Dev. Econs. 9: 121-130.

T y le r ,

Pertanika J. Soc. Sci. & Hum. Vol. 1 No. 1 1993

Export-Led Growth in Malaysian Agriculture: A Var Approach

C. 1973. Exports, foreign capital inflow and econom ic grow th. J. Inter. Econs. 3(4): 337-349.

V o iv o d a s ,

A. 1962. An efficient method of estimating seemingly unrelated regressions and tests of aggregation bias. J. Amer. Statist. Assn. 57: 348368.

Z e lln e r ,

R. 1978. T he role o f ex p o rts and foreign capital in Latin American econom ic growth. Southern Econ.J. 45(2): 410-420.

W illia m s o n ,

Pertanika J. Soc. Sci. & Hum. Vol. 1 No. 1 1993

(Received 11 May 1992)

69

Suggest Documents