Journal of Development
Economics 12 (1983) 219-22.7. IWrth-Holland
EXPORT INSTABILITY, EXPORT GROW
TheBlindersUniwtrsity
of South Awralia,
Pubhshmg Company
1[ AND GDP GROWTTI
Bedford Park, South Austlzrlia 5042, Australia
Received August 1980, final version received May 1,181 The positive correlation
between export instability and expora. expansion, reported by Lam in a of a bias caused by the manner in which the instability index was calculated, and by using data series which end in a, year of unusually nigh exports. Jr is shown that when correction is made for these factors, rlhere ‘q no statistically s~gificant correlation between export instability and exJmrt expansion 01’dome ;tic income expansion.
mnt proper,is shown to be the result
1. Introdactioa
A recent article by Lam (1980, p. 112) suggebits that ‘export instability is most likely to be associated with rising rates of expansion in trade earnings, especially as far as WPCs (Western Pacific Countries) are concerned’. Furthermore, this association, ‘which has not been anticipated in the literature, implies that merchandise fluctuatlonr; may also be correlated to domestic income expansion’ [Lam (1980, 3. 1OS)]. These conclusions are based on the results of statistical analysis which show significant, positive rank correlations between export instability and export expansion for a sample of 15 Western Pacific Countr?ies and 9 Developed Countries for two time periods, 1961-1972 and 1961- 1974. This note demonstrates that the positive correlat’lon between export instability and export expansiogl reported by Lam is the result of a systematic bias in the manner in qwhichthe instability index was calculated. The apparent positive correlation between e,lrport instability and export expansion vanishes once t:orrection is made f(,r this bias. In addition, the alleged positive corrclatior: between export insf:ability shnd domestic illcome expansion is shown to bt=non-existent. 2. Data and methods index used by Lam (1980, p. !0.“1)is *the standard error of estimate normalised by the mean. The star Jard error of estimate was The instability
*I should like to thank W-J. Wallace, P. Morgan, I\!. Polasek and J. Waq’les for useful comments. Final resoonsibilit!! for aq remaining erro-s is I nine.
0304--3878/83/OQO/$OI3,~
g> 1983 North Bolland
calculated by takir g deviations of the value of merchandise exports from their trend values, and the trend values were obtained by a simple linear regression of export vallues against time. Examination of the dalta of merchandise expm values of the c:ountries in the sample rewals that they are in fact non-lkaat with respect to time, Fitting a simple linear trend to non4inear data will ‘give rise to biased regession ’ coefficients since tisspecification o ~’ the regression equation will result in first-order autocorrelation of the residuals which cannot be corrected for in the usual manner. Furthermore, the standard error of estimate derived from fitting a linear trend to no11-linear data will result in countries with higher rates of growth of exports having larger standard errors of estimate. Thus the instability index as calculated by Lam contains a positive bias in that countries with higher rates of growth of exports will have higher instability indexes. The positj ve rank correlation coeficients between export instability and export expansion reported by Lam (1980, p. 108, Table 2) are therefore not slurP1r&&la rirnsa& Correction for this source of bias can be carried out by fitting a curvilinear trend to the export data, The standard error of estimate as well as the instability index should then be lower when compared with the values of these indexes when a linear trend is fitted to the data. The case of Tar van may be used to illustrate this. Fitting a linear trend to Taiwan’s value 01 merchandise exports yields the following equation: X = -- I. IO28+ 0.34% T (all coefis=ients are significant at the 0.05 level), R2 = 0.7440, D W = 0.6”88, SEE = 0.8965 and I, = 0.5900. X = exports, T = time, 0kV ==Durbin-‘Na tson statistic. SEE = Standard Error of Estimate, ! x = InstaSlity Index. The time period covered is 1961-1974. Table A.3 belolv shows that when a curvilinear trend is fitted to the data, the RZ rises Sgnificantly, the DW statistic shows no first-order autocorrelation 0:’ the residuals and table A 1 shows that the instability index IX drops to 0.354. In order to see if this correction for bias in the instability index tieets the results reported k+yLam, curvilinear trends were fitted to the export values of all the countries in his sample, using the same data sources and covering the same time periods. Th\t: results are reported in the next section,
Tables A.1 an i A.;! present the instability index, growth rate of the value of merchandise t:xyorts and the growth rate of G0P for 15 Western Pacific Gantries and <: Developed Countries for the years 1961-1974 and 1961s A 3 and A.4 present the regression equations used to fit port data and then to calculate the instability ~sta~ the data cover oniy the period 5961-1949
since after ;970, there is a break in the ata series on account of the exclusion of data pertaining IO the I=ountry now known as Bangladesh. Also, the equation for Sri Lanka in tabie A.3 was estimated with data for 1961-1972 since using 1961~19% &!a *boorfit on account of increases in exports for 1973 and 1974 reversmg a declining trend in exports during 1961~1972.
Table, 1 Rank
correlation
coeificients
WPC
between expcrrt growth.“. b -DC
R GZ.‘,
0.5107
0.5500
o.cc113’
R k*lp
0.6537”
0.6500
0.3482
R GX.1,
0.0643
0.2500
- 0.003 5
R %.I,
0.0920
0.1667
-0. I165
-_
instability
and
export -
Total sample
1%l--1974
1961-1972
aSour~~e:Tables A.1 and A.2. bWPC = Western Pacific Countries (see table A-l), DC =Developed Countries (see table A.1 I, R = rank correlation coefficiemt, Gx =average mnual growth rate of merlzhandise exports, = index of export instabilit:: (standard error of estimate + mean), 1, =elasticity of growth of merchandise exports [Lam (1980, p.lOS)J & “Significant at 0.65. -
Table 1 shows that when the instability index is calculated on the basis of a curvilinear trend fitted to the value oi merchandise exports, thr: rank correlation soeffrcients between export inst {ability and export expans’ on are no longer statistically significant except when the total :;ample is used. This is probably due to the very large increases in export earnings experiencei by some countries during 1973-1974. The lac K of any c;:brrelation betwee?e export instability and export expansion for the period 1915H972 supports thir;. For purposes of comparison, the elasticit; index of export growth as lnsed 5y .Lam (1980, p. 105) was used in the arjalysis. Table I shows that except for WPCs, the rank correlation coefficient between export instabihty an:j the elasticity index of export growth is not si again, the significant posrtive rank carrel the boom yez.rs of 1973-11974. cor~e~atiol~coef~cients are Sti1tiStiZXlly sigridicant.
Table 2 i Rank 1mrreiation meffickntz ?;C,XW export instability and emmmic IgC3Wth,‘*b -_
WPC
DC
Tota.l $kInpk:
R$5
0,607lC
0.2500
0.5478”
R G,+.I,
0.2964
0.2167
0.3261
0.0500
O.SW
0.0861
0.3333
Q14!B6
m&-mMW-19&I-~374
1961-J 972 R G,Js
-0.0643 RGJt,. I; -a&wrc(?= Tables A.1 and A.2.
bSax~eas for table 1 with t’he foIlowing additions: G,=average annual growth rate of GDP at current p&es, GYp--average annual growth rate of red GDP. “Significant at 0.05.
Table 2 presents data regarding the alleged relationship between export instability and donr,estic income expansion (measured by the rate of growth of GDP). For the period 1961-1974P there is a positive and significant rank correl~i.on c&fkient between export instability and the growth rate of GDP at current prices for WPCs and for the total sample. Howeuer, this positive correlation *ttanisheswhen data pertaining to the 1961-1972 period are used, suggesting that the high prices of exports experienced by some developing countries in 1973-1974 explains the positive rank correla,tion coeflicients observed wflen 1961-1974 data are used. In addition, table 2 shows that there is no rank correlation between export instability and the growth rate of real GDP in either of the two time periods.
The posiive rank correlation coefficients between export instability and export expansion reported by Lam are a statistical mirage caused by a bias in the c&r Jation of the instabili,ty index as a result of fitting a linear trend to non&err data. Correction for this bias shows that there is no significant rank carrel ation between export instability and export expansion particularly for the Ptil-1972 period. In zlddition, the alleged positive correlation * tween e.qort instability and domestic income expansion is also shown to nonexi ;tenh particularly for the 1961-1972 period. Only by extending the to irclude 1973 and 1974, years of unusually high expor xatical analysis be induced tto generate any suggestion of ~~rr~la~i~)~~ between export instability and export expansion or domestic ion.
6. Tan, Export imtahilit~* and pmwth and GDP growth
223
Appendix ’L’ableA 1 Export instability, export growth arld GDP growth, 196t-1974.” -4s
--
Gi
%
%*
South Korea Taiwan Papua New Guinea Indonesia Hong Kong Thailand Singapore Philippines Australia Malaysia New Ze&md India Pakistan Sri Lanka Burma
0.283 0.354 0.710 0.611 0.128 0.277 C-289 Ct.229 (?I50 ? .228 Ii.154 0.13 1 0.047 ‘3.185 0.188
0.434’ 0.234 0.276 0.2lC 0.170 0.150 0.144 0.132 0.132 0.126 0.088 0.085 0.157 0.028 0.003
0.27 1 0.168 0.126 1.731 0.140 0.124 0.136 0.159 0.107 0.117 0.094 0.116 0.093 0.089 0.072
0.089 O.100 0.073 0.045 0.087 0.07 ! c .080 0.053 0.05 1 0.080 3.055 0.031 0.056 0.050 0.03 1
Japan Netherlands Italy West Germany Benelux France Canada USA UK
0.188 0,175 0.119 0.165 0 150 0.143 (I.@95 0 177 1109i
0.210 0.166 0.165 0.162 0.159 0.150 0.138 G.125 6).099
0.167 0.112 0.116 0.089 O.liZ O.jO9 O.;.oO O.(176 0.087
0.08 5 O.OSO 0.047 0.041 0.036 0.052 0.054 0.03 5 0.026
---
-_--
‘Sources: (1) Intematianzd Mmctary Fund, internatrsnal F ‘mm-id various issues. (2j United biations, learbook of National kccounts 1976. (3) United Nati ms, Yearbmk ~j International Trade
Statistics, Statistics, Statistics,
1976.
Table A.2 Export instability, expot I grmth
(and GDP growth, l%i-1972.”
224 Table A.2 (continud)
!+fewzedand
1,
‘Gx
%
%*
0.105
0.083
0.175
d 0,039
0.084 0.132 0,119 0,101 0.093 0.067 0,067 O.~Sl
0.083 0.050 0.078 01032 0.056 * 0.053 0.042 0.033
0.159 0.102 0.088 O.lO9 0.104 0.092 Q.087 0.072 0.078
0.092 0.046 a033 0.050 0.053 0,043 0.054 OS037 0.026
u7i 0,076 .a@72 0.062 0,055 W69 o&v 0.125 0.047 m6o . 0.074 -0,008
F hilippines S ingapme Iijdia F’akistan Malaysb rSri Lanka I LUlllMS
t-I.054 0.044 0.073 0.054 OAI65 0.031 QO66 O*OZ 0.045
Japan
Jiealy IkUdUX
?-Jetherlands I?ae.ce ‘Vat Gmmny Canada IfSA FJK ‘Sources: Same as in table
0.191
0,146 0.141 0.133 0.126 0.125 0.121 0.080 0.076
A.1. Table A.3
Rcgressio:1 equations
X=a+b,Tf?#,
a
1961-1974 (t&tires in parentheses errorQ b,
ba
R2
are standard
DW
p
1.01
0.4939
WPC
South K
Qa643h XI.9649
1.8423b (0.2471)
- 0.6520b (0.3270)
(o;oa74)
Taiwan
2:21Mb (0.2498)
- 0.‘r3944 (0.1477)
0.0690b (0.0082)
0.9616
1.14
0.4284
Papua NW Guinea
1.6371d (0.8476)
- 0.6936’” (0.2577)
0.0632b (0.0169)
0.7488
1.59
0.2045
In&nest 1
2.2311” (0.8650)
*-0,7860b (0.2653)
0.071SC 0.7642 (0.0172)
1.56
C 2158
Hong Kmg
l.183gb (0.2659)
- 0.254Sb (0.0815)
0.0406b (0.0053)
0.9716
1.18
0.4076
7lklanc
0.8101b (0.2213)
-0.1641” (0.0679)
0.0176b 0.8249 (O.oi)ctl)
1.17
0.4128
2.030sb (00.48f3)
-0.519Sb (0.1492)
0.0506b ~0.0097)
0.8630
1.32
0.34(i4
0.831@’ (0.2293)
-0.112$ (8.0703)
0.015Ub 0.8276 (0.0046)
1.41
0.2943
3.3027b (0.6272)
- 0.5624” (0.1923)
0.07htb (0.012s)
1.10
0.45 20
0.9403
@. Tun, Export imtahiliey and pvwth and GDP
gqrmvth
225
Table A.3 (continued) ----
-a
b, -_--0.2ri45’ $I. 1084)
b2
R2
DW _w--____I 0.834 I I .30
p 0.3511
0.885 1
1.43
0.2855 0.3594
-
Malaysia
1.568gb (0.3534)
New Zealmd
l.0736b (0.1873)
-0.1:!09”
India
1.7974b (0.2457)
-o.lc~lob (0.0753)
8.019@ 0.8605 gMO4!~)
1.28
Pakistan”
l.6652b (0.0914)
Q.l171b (0.0280)
O.OOtf# 0.9610 (O.OOlS)
2.36
Sn Lanka
0.435gb (0.0181)
-o.0150c (0.0006)
0.0005 (0.0004)
0.8310
2.07
0.1846
Burma
0.3152b (0.0306)
-0.0341b
O.(;fil6b Q.70! 5 (0.~~;;?6)
1.33
0.1884
Bendux
6.8200b (1.4866)
- l.3986b (0.4559)
0 19zb jO.~29i)
0.9562
1.01’
0.4926
Canah
7.3030b (1.2146)
-0.8121’ (0.37 25)
0.1727’ \0.0242)
0.9753
1.43
0.2855
Fran=
X2.0557b (2.2576)
- 2.586db f0.69 24)
0.3359b (0.0449)
0.9601
l.od
0.5224
21.8761 b (4.8975)
-4.7914b (l.c;O.Zl)
0.633gb (0.0974)
1.14
0.4282
Italy
6. 1276b (1.3119)
-0.9or51” (0.40:?4)
0.17iBb (0.0264)
1.46
0.2675
Japan
9.5697” (3.0446)
- 2.7413” (0.933 5)
0.39 1Tb (0.0605)
1.59
0.2053
Netherlands
8.01S6b (1.8357)
- l.9K!Zb (0.56311)
0.2429b (0.0365)
1.18
0.4099
USA
30.1009b (6.5581)
- 4.2412’ (2.0172)
0.51 52b (0.1308)
1.31
0.3450
UK
14.133Sb (1.5718)
- 1 4952b (0.4821)
0.2143’ (0.03 I :;)
1.27
0.0’-)14
(0.0575)
(0.0094)
0.3287b (cl.0070) 0.01 57b
(0.0037)
-0.1800
DC
West
Gemany
aa
_-_.
- _--.--_-
= constant,
b, , b2 = -egression coefficients, R2 = coeffkient of determination, DW = Durbin-Watson statistic, = k&order autocorrelation coefficient, P x = $$alueof merchandise exports, = time. T bSignificant at 0.0 1. “Significant at 0.05. %ignifican~ at 0.10. “1961- 1969 data. fBetween the upper (1.05) and lower (0.713) limits of the DW statistic at the O.U.1 Itz~cl (estimated by extrapolation). However, applicatioia of the Swed-Eisenha; + (19431 Runs Test indicates that there is no autocornzlation in the residuals. Correctifan four first-ordttr autocorrelation of the residuals in SOIIH*of the equations \V~S carried out bq a gerreraliseti difference tra~sfo
,,Tq&eA.4 Re
ion equations, X=re+bJS
i;T 1, 1961-1972 (figures in parenthkses a~ stmdarc! errors)?
1
1.83
0.0872
0.9603
1.24
0.387t
0.9831
2.54
-0.2715
0.9970
2.05
- 0.0274
0.9612
1.89
0.0531
0.7734
1.61
0.1944
-0,1173qb (0.0330)
o,;nSp+ 0.9196 (C&0025)
1.64
a1777
-J&50 (o.o?8(i)
iiMitY 0.8359 (O.OO?!!)
1.63
’ 0.1866
0.9654
2.05
- 0.0277
- 0.0742 (0.&367) -&lor (0.Q - 357) .
(0.1093) o.* (0.042 1)
O*Q461b -0gi.b~ (0.0151) (O.OOll)
1.285Sb @.0770)
-0,1358b (0.0272)
o.01?2b (0.0020)
0.9.c.i1
2.49
- 0.2466
1.542P (0.1306)
-0*0147 (0.0454)
0.0062”
0.8237
2.14
- 0.0700
l.&iib
-@&I71 (0.0209)
2.36
-0,0209 (0.0267)
1.62
0.1890
&m914) 1 MEi?@ (0.0760) o*4359b @Ol81) uma
0.0309b (O.oo(i5) .-. t$$ligE (W@2% ..p:i”
u.2979b (r[).O294)
-U.OEXF fO.rwxw) -0.0216d (0.0104)
s.3449b (0.7079)
- 0.6356” f0.2504)
4.3011b 1’$*425q
0.0083 (0.4485]
4.0518” (Q6141) 5.112Mb
Q.ooo5
0.8310
2.07
0.1846
0.0005 jU.0007)
0.7802
1.37
0.3138
Q.2096b 0,993:, {0.0187)
1.07
0.4642
0.0927b (0*0113)
0.9924
1.87
O&69
- 0.2199
(0.2172)
o,093sb (0.0163)
0.9775
2.13
- cm658
- ~.~~ 5” {0.16@)
O-1148b (U.1148)
0.988:.
0.87f
(O&004)
OS653
C. Tan, Export
instabiliry a .l t~rwdt
227
urrd GDP growth
Tatble A.4 (corltinued) _Io--
DW
P
-.?a
France
8.531&P (0,869G)
- 0.7843” (0.3075)
0.1772b (0.0230)
0.9813
1.33
0.3312
We ,t Gernrany
13.9158b (0.7798)
- 0.7584” (0.2758)
0.2;fzO!’ (0.0206)
c-3956
1.23
0.3834
Canada
3.178Sb (0.5887)
0.043 5” (0.0183)
0.9828
2.15
- 0.0762
0.1 222b (0.0224)
0.9929
2.71
- 0.3533
0.1227’ (0.02 11)
0.9687
2.13
- ).066C,
0.83Zb (0.226’:“)
0 953[jb (0.298’?) ._-Q*5:‘9<“~ I2 7490b UK {!?17969) (0.28 15) ----_ --.P “See table A.3 for key to symbolis and footnotes.
USA
19.792:8b iOM.r:‘O)
._P.---
References Lam, N.V., 1380, Export instabilllty, expansion and market concentratio.1: A metlwdological interpretation, Journal of Developm,nt Ewnomics 7..99-l 15. Swed, F.S. and C. Eisenhart, 194:;, Tables for testing randomness of grouping in a sequence of alternatives, Annals of Mathematical Statistics 14.