Journal of DcvclopmcntEconomics 12 ( 1983) 229-236. North-Holland Publishing Company
INSTABIELITYAND THE GROWTH OF EXPORTS A Misinterpretation of the Evidence from the Wesiern Pacific, Cotmtries Constantine GLEZAKOS* CaljfornkaState University,Long Beach, CA 90840, USA
kceived May 1981, fmal version r,xeived February 1982 This paper points out the methodological weak lesses in NV. Lam’s article axl w,ith the employment of approptiate measures demonstrates the tautologcal nature of Lam‘\; findings on the relationship between export growth and instability. In addition it provides certain general guidelines regarding the sekction of export instability indices.
In a recent article in this Journal, NV. Lam (1980) claims that by investigating the question of export instability ‘from a differer.t methodological basis’ and by avoiding the ‘important specification problems’ which were ‘the source of inconsistent results and possibly misleading interpretations’ l in prior studies, he has proven that ‘export instability is most likely to be associated with rising rates of expansion in trad(i: earnqs, especially as far as WPCs are concerned.‘Z The objective of the present Note is primarily to demonstrate that Lam’s statistical results, especially those indicati;lg the Gxistenze of a positive relationship between export instability aud export growth, are tautological and that some of his conclusions are urwarranted. The single most important methodological weakness of the article is its measure of export instability. The author has, of course, given two reasons for choosing the particular instability index employed, namely (I) that ‘:here exists convincing evi.ience that the selection of any one of the available instability indicators would not affect comparative results for a crosssectional study within the same period’, and (2) that his index ‘has been one of the two most frequently used indices’.3 As will be demonstrated below and *The author is thankful to Professor Jeflrev. Nugent for cortsfructive comments 011 an (Earlier version of this paper. l Lam (1980, p. 99). 2Larra (1980, p. 112). %am (1980, p. 105).
030&3878/83/
/$03.00 Fi> KU33r\torth-Holland
as has already been shown by several authors however, the choice of instability index does figure importantly ir, the results of export instability :Ptudies.4More signifkmtly, the above rationalization regarding the choice of tte export instability index manifest ESthe author’s failure to take into consideration the kndamental principle behind the measurement of export instability. ISecause~ and as’ common sense suggests, an export instability index cannot tpe a fortuitous measure (or even a measure based on popularity), but one that is congrknt to ‘the ti priori assirm@tions of the efkts of export instability that the particular study attempts to test. Particularly, for d study liI& the 6ne under consideration whose objective is zhe determination of the: r&e of export instability on the growth of exports, a proper export instability index should at least be independent of the trend (or growth) rate of exprts s export ins?ability has been defined explicitly or In the relevant Ilterature, _’ implicitl;l as a measure of the ‘unexpected’ or ‘unpredictable’ changes in export revenues. Consequg:ntly, most of the known export instability indices are designed to measure, in one way or another, the ‘uncertainty’ rather than the ‘shortfall’ effecE:of foreign earnings. 6 It is in this context tk at Masseil, among others, has correctly suggeskd that ‘in constructing an export instability index, it is necessary first t J eliminate the trend. Otherwise, a country whose exports are growing rapidly - even at a constant rate - will score high on the instabikrty scale.” In other words, as long as export instability is cq3nsidered as a measure of the degree of uncertainty in export proceeds, it should be obvious that for a country whose exports grow at a constant rate (or cflnstant increments), export instability should be zero regardless of the magnitude of the growth rate (or size of the increments) of exports. This is because in such instances there would be no ‘unexpected changes in exports and thus no perceptible unckk~ainty regarding export pro~~Ms. From the above it follows that the choice of the trend to be used for the estimation of the residual variability in export earnings cannot be made arbitrarily. Rather, it should be dictated by the actual performance of each country’s time series of cxporlts. In the light of the preceedingdiscussion it is obvious that Lam is guilty of two conceptual errors. First, he failed to difTerentiate between the ‘shortfall’ an ‘uncertainty’ asp&s of export instability, as can be deduced from his ‘For
example, Lawson (1974, p. 55), and Eeith (1970) arrive at the same ccnclusion. liy the iatter states that ‘the conclusion one draws regarding the change in kstability fad on one’s definition of instability and hence on the index employed’ (pp. 267-268). quotes Leith as evidence in support of his claim that the choice of index is Lam’s footnote 10.) e, the index should be desigtled so as to be in harmony with the tested reasons for which export instability influences export growth. Nugent (1976, p. 330).
criticism of the trend correction prior to measuring export instability because, according to Lam, such rjractice implies that ‘all abt;olute changes, whe:ther in the Sante or opposite direction, . . . are of equal significance? Second, but more importantiy, he confused export instability with export growth as it can be easily inferred from his disapproval of the Jetrended export instability measure because it ‘implicitly assumes that a changes [in exports] . . . and all constant rates, whether at kgher or lower speeds, are of equal significance? This latter misinterpretatkn of instability has a significant bearing on the outcome of the study sirce it Ihas affected the author’s choice of the export instability indicator. Lam defined as his instability index the ‘normaliz:d standard error of estimate of the ‘linear regression of exports against tima:‘? The atioption of the linear trend as the appropriate form of export growth for the estimation of export instaoility was arbitrary; no attempt whatsoever was made to use eve:t a sznple ‘goodness-of-fit’ criterion as is indicated for in>;iabil,ty indices measuring ‘uncertainty’. ’ ’ Furthermore, it can be easily shown that if a country’s exports grow at a constant rate, i.e., if ;Acountr;& e:Kports (X) can br: represented by (1): X,=X1(1 +x)‘-l,
t=
1.2$..., n,
(11
where x is the annual growth rate and X, the initial value of the f :ries, then Lam’s e:.port instabilify index’ (I,) and ‘elasticity index of growth’ (E,) can be both reduced to expressions that are illcreasing functions of the growth rate (x) and of the sample size (la).l2 This on the one hand means that for countries whose exports grow at perfectly stable rates over a given period of time., those with higher growth rates would show greater instabi.lity according to Lam’s indicator while, in reality, noale of them experiences any instability. *Lam (1980 - , p. 102). It is worth pointing out that despite his repeated criticism of past studies, Lam himse”rfemployed an export instability indicator which both corrects for the trend and does not differentiate beltween absolute changes in export earnings of the sttme or opposite directions ‘Lam (1980, p. 102). Lam’s co,rfusion of the conxpt of export ins itbii;:y wltn that of export growth is also evident in several other statements like: ‘it is the stabi; TVc f expclrt receipts which should be a cause for greater concern’ (p. 109). Here again by having identified instability as fast growth, Lam cannot perceive that stability in export earning! tl )es not necessarily impi) stagnation and that it is feasible for stability in exports to coerist v*;th either slow or rapid growth, as Inng as such growth takes place at constant rates. “Lam (1980, pp. 104405). “See e g Cuddy and Della kalle (1978) for a concise yet informati\c discussion ox this ropic. ’ ’ rhlse’ eipressions are:
On the other hand,, given that &, is also higher fur the countries with more
rapidly growing exports, it f&lows that the high correlation, that would be extibitgxi .b@wcg~+ nthests.’two mea8urem43nts:i for kcountties c with steadily ’ : 2” wml
of. the *countries- ~inchided~ 1ifi
Imxf’s
1sample, ’
-
n0tabXyHong KOngl
Karea and Taiwan, ~~x~~en~~~~~a~~dgrowth in their a,exports*(16% to *39x for 1961 to 19’7Zor 18%,to 41% hr :I951 to &9’74$Furthermore, E&three of thm ’ amtries enjoyed, relatMy stable’ and +zraoth export growth as is inti by the high, determination coe&~e~& (over 0.95) in the fitted exponential trend lines (I), a fact which should rank them lo\v on the ~nstabiljty scale. 13 Yet, according tc the mesure of export instability employed by Lam tlhey are ra&ed at the top of the countties in his sample, This suggests the inordinate degree tea which the .statistical results on which Lam based his conclusians ,folfow tautologicafty from the unfortunate but seriolus flaw of sy&zzratic rneasu~~ment error, in hk instrtbility indicator. To demonstrate that this is actually the case I have estimated for the period 1961 to 1972 (as well as for 1961 to 1934) Lazn% export instability index for the 14 WPCs using however both the linear and exponential trend lines. These two indices, whose country specific values are presented in the statistical appendix, are: --rl’“-(~~/~)~~u:)2/(n-2), where X,*=(P:+& for the lirzesr trend, and I~e=(tOOIX!~(X,-~~)2/(n-z), where d, = A(1 -I-xr for the exponential (log-linear) trend, and x is the sample average. In the ziubquent statistical testing, the fitted growth rate (x) estimated frc;nr (I) ‘~r’%s: used as an additional measure of export growth. This was done in order to avoid the possible bias that might be involved by representing export growth with Lam’s ‘elasticity index of growth’ (E,) and because it is a ore appropriate and generally accepted measure of growth.14 Tabfz I ft,‘fow presents the pairwise correlation cofkients between the two easures lof export growth (x, E,) and the r;wo measures of export instability, i.e.. Lam’s I!p and the index estimated from the trend with the best fit for
is ~x~~ain~ by Cuddy and Delia Vaile (1978, p. 82), ‘the better the “fit” of the wer I, since there is less variation about the regrewon model to be as in E,. (1) the poor fit of the linear trend for countries existence of an upper bound.
each country (lp). i 5 As can be SCen from table 1, for berth time periods, there is no significant correlation ?+etween export growth (u or E,) and the export instability index est&ktP Ir”J from the trend with the best fit (ii”“‘). On the contrary, and as was expected from the foregoing analysis, Lam’s instability index is indeed posWely and significantly correlated with both
export growth measures.
Table 1 Pairwise correlations between export growth and initahilitq. Export instabi!;ity index.’ fin x
eX;
pCS’ x
L
-
1%1-i9?2 Export growth rate (x) Elasticity of growth (E,)
0.793b 0.747b
-0.006 0.03c
- 0.224 - 0.242
0.7Mb 0.732b
0.156 0.168
0.016 0.025
i HI-1974
Export growth rate (x) Elasticity of growth (EJ ‘Source : Statistical appendix. bSignifican t a t the 1% level.
In order to further demonstrate that the non-significant correlation between expor; growth and I:est was not just accidental, a totalq d.ifferent expomr;instability measure was employed. This index is based on adaptive expectations and is estimated according to (4)
where Xr Idenotes the level of export revenues expected at year defined asl”
x:=x;_
1
i-&q_
1
-x;_
1).
r
and is
15)
“It should bt minted out here that the ‘goodness-of-fit’ criterion used for the choice between coefficient. For a meaningful log-linear trends was their determination the 1near B ever, the ciete~inat~o~ coefficient fur the log-linear trend \\aas obaained by: coq.mrison, RZ = 1 - /&X, - ST);“,where e, = X, - S, = X, - A[1 e x)‘. ’ 6F or brevity reasons, ?hL4derail * regarding I e wrch pfi~ct.“dursused for the es:imation of the aJjustmenii coeffkknt /i and of he expected exports have been omitted. However. the gener;tl methodology can be found in Glezakos (1978).
r
This ind,ex,*whosesample v&es tire prr;seated in the statistical appendix, ‘is a measur@lof the unpiedictabfe Swings i3nexports, since: it aflaws the systematic export dhanges to be anticipated. 17;It is uot surprising, of course, that 8s is shown in the last column of table 1, neither IyP is significantly correlated ‘th &her ‘measure ;of -exports-gtowth~-forlibaththea :priods ~on~id~~‘?Yhe latter tiding lends further support to the claim rriade‘ earlier about ‘the tautological nature of the significant raak correlation between 1:” and E, obtained by Lam. We admittedlv, there exist * reasons for expecting either positivle or negative effects of ixport instabmii 3~ “&xport&roWh, Lam ‘neither presented any consistent explanation as to why- export instability should be a stimulus to export growth, nor djd he use a suitable testing procedure that would have enabled him to confirm the positive association, On th,e contrary, the composition of the sample which includes a non-homogeneous group of countries’* provides a very plausible reason for which, even the employmen; of more apprqriatc export instability measures should not show any signG&zantrelati~~hship with export growth for La&s sample. Iln conclusion, the following remarks are in order, The findings preserted in this Note should leave little doubt that @he positive association between export instability and export growth sholvn in Lam’s article ;is primarily the result of the tautological nature of the instruments enqployed for the measurement of these variables. Consequently, the author’s repeated claim that he has discovered a relationship that ‘has not been anticipated in the relevc.nt literature’1g is prec@tate. Although the present Note hits esclusively concentrated on the export growth and instability aspect’ rtif Lam’s article, it should not be difficult to infer that the shortcomings of its export instability index render some of the other estimated relationships dubious, especially those involving the ‘elasticity index of market diversification’ (D&. This is because DeXis also based on the estimated linear trend of each country’s annual export shares ‘outside the traditional six most important partners’.20 Finally, it should be abundantly clear from the rationalization as well .QS from the results presented in this Note that, Lam’s clarm “X1 can fx easily shown that Iyp would be zero for a country whose exports change at a mtaot rate durmg the period under consjderation, regardless of the size of the annual growth countties in the sample two are developed (Australia and New Zealand) m oil producer (Indonesia) and as has been shown elsewhere expcrr! instability does ~~~~i~ca~t~y export growth in the DCs. For detaiEssee Glezakos (1973, especially pp.
and I 12).
C. Glazakos, irrstahilit_y and th p-m
tlz
cf exports
235
notwithstanding, the selection t>f the export instability index does indeeaf make a considerable e results of cross-se(ctional studies or\ the aSects of export instability.
StatistkxCqpendix Table A. 1.” _-
-_I 1961-9972 -.
--
--
Export giY?WUl
Country
rate x
1119 x
1’09 x
I:“P
1I,_$
Austn alia Burma Hong Kong India Indones ,a Korea Malaysia New Zealand Pakistan F’hilippines Singapore Sri Lauka Taiwan Thailand Totals
9.18 - 8.5,” 15.92 4.15 6.82 39.10 5.19 5.89 5.67 6.77 6.32 - 1.57 26.02 6.64 127.55
12.67 17.77” 14.81 7.71 24.71 45.76 6.72 12.92 9.01 b 6.9Sb 16.32 5.43b 44.24 11.80 238.82
8.26’ 18.00 3.64b 7.215 25.21’ 8.Wb 6.21b 11.9,ab 10.39 7.32 14.62b 5.48 21.83b 10.88b 159.019
6.66 12.84 3.74 6.25 12.28 7.31 5.96 9.74 7.47 7.77 9.48 4.80 8.99 8.92 112.21
59.03 - 57.18 93.17 27.1 ! 50.7rj 169.83 33.24 38.79 33.79 40.92 43.23 - 10.25 143.52 42.48 708.4 1
-
Coiultry
1961-1974 ___I__ Export growth rate x r!jn
Aus*.ralia Burma Hong Kong Inclia IncYonesia Korea Miilaysia New Zealand Pa,kistan Philippines Singapore SIYLanka Taiwan 1 hailand Totals
11.83 -- 5.71 17.68 6.3 ! 14.5Q 41 .jJ (.8’{ :2.G .81 13.07 11.35 0 48 28 h5 10.5 I 171.21
---p28.94 2 4.02 31.26 20.28 94.70” 84.34 35.33 24.42 13.55b 35.13 52.63 14.53b 59.13 41.73 559.99
-
-
111% x
exp IX
22.02b lW0 22.91” 13.87 15.66”’ ?. ‘0 18.40”’ 8.41 95.13 38.23 29.83b 12.27 32.53b 14.49 21.37b 13.01 14.86 7.69 31.12” 1r .96 49.13ib 11d.88 14.‘58 5.13 ‘7..j3h 8.69 37.1$8b 14.18 432.‘75 188.21
EX 90.56 - c15.69 119.89 50.93 140.98 203.04 74.71 66.37 49.69 80.27 99.88 5.33 172.0-t
87.63 1195.63 -aSource: IMF lvater’nationcrlFinmacial S!ntistic s (variou> issues). bDenotes index obtained from trend with best-fit on the basis of R2. (See footnote IS abcbve.b
Referedkes Cuddy, 3.D.A and I?& D&a hb, 1978, Measuring the instability of time series data, Bulletin Oxford UGversity institute of Economics: and StaGstics40, Feb., 79-85. Glezorlros,C., 1973, Export instability and economic growth: A statistical verification, Economif: D~eiopm~~nt and C&urak Change 21, July, 670-678. Gfezak~os,C., ~,978,,fnflationand growth: A reconsideration of the evidence~fromthe LDCs, Journal tbf Develc$np Areas K&Jan., 171482, brn, N.V., 19110,!I?xport iastabifity, expansion - and market concenttation: A methodological imerpretation,Jcutmaiof DevelopmentEconomics 7, March, 90-l i 5. Lawm~ C.W., 1974, The decline in world export &stability: A reappraisal, Bulletin Oxford University Institute of EWnomics and Statistics 36, Feb., 53-65. L&h, IIC, 1970, The decline in wozrldexport instability:A amment, Bulletin Oxford University Institute of E?zonor&s and Statistics 32, Aug,, 267-272. l%sseNl, B.F., 1970, Export instability a& e;conomic structure, American Economic Review 60, Sept., 616-630. Y~~topr~uIo~P.A. snd .J.B. Nuent, 1976, Economics of dtivdopment: Empirical investigations (Warperand ROW,New York).