Export Growth, Economic Growth, and Development Levels: An Empirical Analysis

Jessie Poon

Export Growth, Economic Growth, and Development Levels: An Empirical Analysis

The export-promotion development strategy has been advocated as a generally valid prescription for economic growth on the basis of econometric studies of the exports-growth relationships. A prominent shortcoming of these studies is in the presupposition, intrinsic to them, that exports-growth relationships hold similarly fn- all developing countries. This paper tests the hypothesis that the exports-growth relationships vary with the developing countries’ level of development. M y results indicate that high export growth is most beneficial to developing countries at intermediate stages of their development cycle, and consequently suggest that the export-promotion development strategy is especially appropriate only for these countries. 1. INTRODUCTION

The role of exports on developing countries’ economic growth has received considerable attention since the 1960s. Exports have been viewed by some authors as an engine of growth (Ftiedel 1987). Others have credited them with contributing to the reproduction of uneven development patterns, on the presupposition that developing countries are predominantly engaged in low-value primary exports subject to deteriorating terms of trade (Singer 1984) and to a low price elasticity (Prebisch 1962). Several empirical studies focussed upon the effect of exports on economic growth. Many of them tended to side with Riedel (Table 1) with the result that developing countries have been strongly encouraged to pursue export-promotion (for example, World Bank 1987). However, there have been considerable differences in the estimates of the effect of exports on growth, namely, on the reported size and significance of the slope coefficients associated with export variables. Only recently some researchers questioned the failure of these studies to differentiate among heterogenous groups of developing countries (Helleiner 1986). This paper argues that the different findings produced by the studies on the effect of exports on growth reflect not “noise” or “sampling errors” but contextual parametric drift. Thus, it is unwarranted to generalize the exports-growth relationships a-contextually, by assuming that they hold with the same parameter values across all developing countries. The conventional cross-sectional studies based on The author gratehlly acknowledges the comments of four anonymous reviewers.

Jessie Poon is visiting assistant professor of geography at the University of Georgia. Geographical Analysis, Vol. 26, No. 1 (January 1994) 0 1994 Ohio State University Press

38

/ Geographical Analysis

untested presuppositions of parametric invariance are bound to produce little consensus on the contribution of exports to economic growth. Instead, we should search for contexts across which the contribution of exports to economic growth will differ. Namely, we should ask: for which contexts is the contribution of exports to growth greater for countries at some points of a context, than for countries at other points of the context? The context selected for consideration in this study is development level. This draws from the observation that the export coefficient estimates tend to be larger for higher- than lower-income developing countries (Michaely 1977; Tyler 1981). This study implements the Expansion Method (Casetti 1972, 1986). As a research philosophy the Expansion Method repudiates the assumptions of parametric stability on which the conventional model specification and model construction are based, and emphasizes instead the need to link models to their contexts (Casetti 1986; Taaffe and Casetti 1990). To this effect, it suggests that the potential drift of theoretically grounded models across relevant contexts should be searched for, theorized, and tested. This approach is especially relevant here. The unqualified advocacy of export-promotion as the appropriate development strategy for all developing countries is grounded on unwarranted extrapolations from the econometric estimates of the exports-growth relationships. In fact, the literature is replete with disagreements over the relative contribution of exports to growth.2 Many of these disagreements may be attributed to the presupposition that the exports-growth relationships are invariant when they actually are not so. I contend that the exports-growth relationship is context-specific, and specifically that it varies with a country's economic development stage. This in turn implies that export-promotion is not equally effective at all stages of the development cycle. The paper is organized as follows: the theoretical and empirical literature on the relations between exports and growth are addressed in section 2. Section 2 also documents why the exports-growth relationship is likely to drift with development levels. The Dual Expansions are introduced in section 3. The empirical analysis is discussed in section 4 and the final section concludes with the implications of the empirical findings.

2. LITERATURE AND THEORY

The export-oriented development policies have been advocated on the basis of market principles, dynamic learning, and technology acquisition considerations, the need to release balance-of-payments constraints, and to achieve employment and income distribution gains. It has been argued that exports make it possible for developing countries to overcome the limitations of their domestic markets and to attain greater capacity utilization and scale economies (Balassa 1988). Also, it has been maintained that export-oriented development policies produce a more pronounced industrial progress (Joint ECLA/UNIDO Industrial Development Division 1986; Westphal 1978), lend themselves to importing industrial capital (Spetter 1970), and bring about higher-quality products because of the exporters' exposure to international consumption patterns (Krueger 1985). 'Export-promotion refers to the ado tion of a structure of incentives that does not discriminate against exports in favor of the home maget (James, Naya, and Meier 1989). 'Relatively strong support for the promotion of exports can be found in Balassa 1978a, 1978b, 1985; Bhagwati 1978, 1988; Krueger 1978, 1985;Little, Scitovsky, and Scott 1970; Ranis 1985; World Bank 1987. For more pessimistic views expressed toward the promotion of ex rts, see Adelman and Robinson 1989; Adelman 1984; Cline 1982; Lewis 1980; Nurkse 1961; P r e g c h 1962, 1984; Singer 1950, 1984, 1991; Singer and Gray 1988.

Jessie Poon

/ 39

Developing countries are commonly short on capital goods that have to be imported. Their import capacities are dependent on their export earnings which provide the necessary financial means to achieve the material prerequisites for industrial development. Finally, Krueger (1988) argued that an export-oriented approach in a labor surplus economy permits the rapid expansion of employment and real wages. Since developing countries tend to be more labor abundant, export-promotion encourages the use of labor and the growth of employment. Banerji and Riedel’s (1980) study, for instance, indicates that Taiwan’s rapid growth in industrial employment has been enhanced by its shift towards laborintensive export activities. Balassa (1978a, 1985) suggested that income growth is achieved at a substantially lower cost by investing in export-oriented countries. The efficiency in the export sector has been discussed by Feder (1982) who argued that the export sector generates growth by increasing the aggregate levels of labor and capital. Bhagwati (1988, p. 5) concludes that even if exports do not contribute directly to total factor productivity, a more efficient regime will require less savings to sustain it than a less efficient one. Since the 1960s, a number of empirical studies have focussed upon the exportsgrowth relationship. Some of these studies regress indices of economic growth versus export and/or export growth indices; others employ a production function framework. Table 1 summarizes some of the exports-growth analyses. In one of the earliest studies, Emery (1967) obtained a highly significant coefficient for the export variable indicating that circa 3 percent export growth is associated with 1 percent increase in GNP per capita. Similar results were obtained by Syron and Walsh (1968) and Maizels (1968). The more recent studies tended to employ a production function frame of reference, and possibly are characterized by a greater variety of findings. The estimates of the parameters associated with export variables range between 0.04 (Michalopoulos and Jay 1973) and 0.42 (Feder 1982). Their statistical significance varies (see Table 2). However, in many/most cases they are significant at least at the 5 percent level, leading Balassa (1985a) to state that irrespective of the procedure applied, exports contribute substantially to the economic growth of developing countries. A development strategy based on export-promotion evolved over the last three decades drawing from such conclusions. The World Bank (1987) has been especially enthusiastic in its encouragement of export-promotion policies for developing countries. However, if we look closely enough at the empirical studies on record, we notice that their results are quite contradictory. For instance, the effects of exports on growth are more positive for some developing countries than others, and in a few cases are even negative. The tremendous parametric variation in these studies has been recognized by several researchers (cf. Feder 1982, p. 64; Moschos 1989, p. 99; Ram 1987, p. 52; Rana 1985, p. 15; and Salvatore and Hatcher 1991, p. 22). Yet, a research shift aimed at explaining this parametric variation has not materialized. Specifically, the recognition of parametric variation has not led to investigating across which contexts the exports-growth relationship can be expected to vary, how, and why. The shift from estimating relationships to estimating the contextual variation of relationships advocated by the expansion methodology represents a clear departure from the conventional econometric estimation of relationships under an implicit presupposition of invariance. A perusal of Table 2 suggests that the exports-growth relationship is generally stronger for higher-income countries, implying that the degree of positive association between exports and growth depends on a country’s level of economic development. In this paper the variation relationship of export growth to economic

40 /

Geographical Analysis

TABLE 1 Summary of Various Exports-Growth Model Formulations Equation

y=a+bx y=a+bx

+ bkf + cl + dx AY = ak, + bkf + c ( A L ) + d(AX) AY’ = ak, + bkf + c ( A L ) + & A X ) + eY AY’ = ak, + bk + c ( A L ) + & A X ) + eY ( f i x y i = a + bk + cl + dx y z = a + bk + c l + dmx y = a + bl + c ( l / Y ) + dx(X/Y) y i = a + bk + cl + dx yz = a + bk + cl + d,x + d,mx y = a + b ( Z / Y ) + cl + dx(X/Y) AY = a + b(l/Y) + c ( A X ) y = a + bk + cl + dx y = ak,

Description

Source

R2

ICs and DCs 1953-63 ICs only DCs only 1953-63 1960-69

0.67

Emery 1967’

0.86 0.62

Syron and Walsh 1968

0.71

1960-73

0.77

Michalopoulos and Jay 19732 Balassa 1978a

1973-79

0.338

Balassa 19853

0.375 1966-77

0.685 0.714 0.689

Tyler 19814

1970-78

0.57 0.58

Kavoussi 19846

1965-73 1974-82 1960-73 1973-81 African DCs Non-African DCs

0.63 0.20 0.826 0.783 0.303 0.439

Rana 1985’

1964-73

Feder 198Z5

Rana 19888

F~~~19909

Pwms: ICs and DCs are industrialized and developing countries, respectively. 2y = GNP per capita growth rate, x = export growth rate. foreign capital growth, y = output growth, k,l = domestic capital growth, k

I = labor force growth, x = export growth. 3AY, A L , A X = changes in GNP, labor force, merciaLdise exports, respectively, between initial and terminal year as a perUenta e of initial ear value, k,, = sum of gross domestic investments less current account balances fmm initial to terminal 03year. fr = sum o!curre”t account balances fmm initial to terminal year (note: k , and k, are expressed as percentages of initial rear GNP), Y = initial year GNP per capita: W,LX = initial share of manufactured ocds in total exports. 5 y = CNP gmwth, k, 1, x, nu = capital ormation, labor, export and manufacturejexport growth. y = GDP growth, I = labor force gmwth, I/Y = investment output ratio, d = (6/1 6) F, where 6 denotes the Ftersectoral mar inal factor productivities, and F, is the externalities effect of exports on nonexports output. Similar to Tyler5 except mx = product of the share of manufactured goods in total merchant exports. ’Similar to Feder’s eauation. ‘AY = change in GNk, I/Y = sum of gross domestic investments, A X = change in merchandise exports (note: I/Y and A X gre fyxn initial to terminal year as a percentage of initial year value. Similar to Tyler’s. Pooled 19M)-70 and 1970-80.

+ +

growth (exports-growth, for short) will be investigated with respect to development levels. No implication is intended that the variation of the relationship across other potentially relevant contexts is not worthy of investigation. The theoretical bases of the drift of the exports-growth relationship with development levels is addressed next. Economies at different development levels have distinctive production and structural characteristics and dissimilar productive capabilities (Syrquin 1988). Higher development levels are accompanied by a transformation of the productive sectors and by an efficient reallocation of factors from low- to high-productivity sectors. In most cases, sustained economic growth occurs as an export-led “spurt” (Reynolds 1983), so that the exports may be credited with growth effects that are partly due to other transformations of the economy. At any rate, a high growth rate of exports alone does not assure growth (Bruton 1989, p. 1616). Kavoussi (1984) states that it is only when export expansion is accompanied by a rapid growth of resources with major gains in factor productivity that the effects of exports on growth are strongest. This occurs especially during a country’s transition from low to high development levels which is characterized by an acceleration in the growth of labor and capital (Syrquin 1988). Thus, middleincome countries tend to grow faster than lower- or higher-income countries (Syrquin 1986).

Jessie Poon

/ 41

TABLE 2 Various Estimations of the Export Coefficient Manufactured

n

Exwrts

Emery 1967

50

0.67

ICs and DCs

Syron and Walsh 1968

35

0.62

Low-income DCs

0.86

Middle-income DCs

39

0.3295 (9.937)** * 0.3327 (7.443)*** 0.3718 (8.299)*** 0.04 (4.82)***

0.71

DCs

10

0.04 (3.57)**

0.77

43

0.182 (2.457)**

0.30

Semi-industrialized DCs DCs

41

0.57 (1.694)**

0.69

37 37

0.055 (1.604)*

Exports

A2

Remarks

~~

13 Michalopoulos and Jay 1973 Balassa 1978a Balassa 1985 Tyler 1981

Feder 1982 Kavoussi 1984

Fosu 1990

35 31

0.045 (2.23)**

0.031 (1.445)*

73

0.422 (5.454)*** 0.105 (3.72)***

73

0.083 (2.33)**

37 37

0.077 (2.47)** 0.093 (2.48)**

36 36 56 72

0.163 (3.00)*** -0.072 (1.01) 0.123 (2.66)** 0.149 (4.67)***

0.71 0.71 0.75 0.69 0.57

0.00061 (0.97)

- 0.0005 (0.75)

Middle-income including OPEC countries Middle-income excluding OPEC countries Semi-industrialized DCs All DCs

0.58 0.57 0.58

0.65 0.00496 (4.19)*** 0.78 0.30 0.44

Low-income DCs Middle-income DCs African DCs Non-African DCs

NOTES:1Cs and DCs are industrialized and developing countries. ***, **, * Significant at 1, 5, and 10 percent. t-values are in parentheses.

This suggests that some level of development may be necessary before exports are likely to benefit a country’s growth (Michaely 1977; Tyler 1981). The staple theory of export-led growth points in the same direction. In terms of this theory, new or underdeveloped regions or countries develop as a consequence of the growth of resource-rich exports (Lewis 1969; North 1961). The processes involved take place in stages: at the beginning by promoting infrastructural expansion and demand for domestically produced inputs; next, by encouraging the processing of raw material; and finally, through import substitution via the domestic production of consumer goods (Balassa 1989, p. 1665). The empirical evidence has also lent some support to these notions. It has been noted that the export variables tend to have a smaller effect on growth at lower development levels (Kavoussi 1984; Syron and Walsh 1968; Tyler 1981). Kavoussi (1984) further remarks that the correlation between manufacturing exports and economic growth changes from negative to positive as the countries’ incomes increase. Reynolds (1983) intimated that in the typical growth scenario, once a country’s sustained growth is initiated, an acceleration in growth is due mainly to export growth. Subsequently, the rate of economic development levels off as the ratio of exports to national product stabilizes. Two reasons for this sequence are offered. Over time, technological improvements become more critical in lubricating the growth “engine” (Lewis 1979; Joint ECLA/UNIDO 1986). A second reason is that

42

/ Geographical Analysis

in the long run, the once-and-for-all efficiency gains from factor allocation to the export sector cannot supplant productivity gains as the main engine of growth. In Packs (1988) view, further income gains must come from growing productivity. The positive effect of exports on growth cannot be assumed to last indefinitely. The previous section discusses a number of theoretical positions suggesting that the exports-growth relationship drifts with development level. In terms of the expansion methodology, if the exports-growth relationship drifts with development levels, a dual relationship between economic growth and levels of development drifts with export performance. These primal-dual formulations are discussed later in the paper; here I elaborate on the theoretical underpinnings of the dual. One important feature of developing countries’ postwar development is their high rate of growth by historical standards (Morawetz 1977). This accelerated growth has been most conspicuous for middle-income countries embarking on extensive industrialization (Syrquin 1988). This feature is related to the “latecomers” thesis, suggesting that latecomers may grow faster because of the larger body of technological knowledge available to them. Gerschenkron (1962), upon examining the rate of industrial development in relation to the degree of economic backwardness, suggested that the more delayed the industrial development of a country, the greater is the “spurt” of its industrialization. On the other hand, developing countries closer to the low and to the high ends of the development scale are known for their smaller rates of economic growth. Due to shortages of capital and skilled manpower the least-developed developing countries tend to experience very low growth often alternated with periods of decline. And the most-developed developing countries move eventually closer to the lower rates of economic growth of the mature countries, as their catching-up phase ends and their economies become increasingly dominated by the tertiary sector. Empirical analyses and theoretical considerations indicate that the relationship between economic growth and development levels takes the form of a parabola with a maximum, as suggested by Russett, Deutsch, and Lasswell (1964). I contend here that this parabola changes shape in response to lower/higher rates of growth of exports.

3. THE MODEL

The Expansion Method (Casetti 1972, 1982, 1986; Casetti and Jones 1992) represents a departure from conventional modeling, in that it focusses upon the linkage between the models and their contexts. The Expansion Method is a research philosophy and a technique for constructing and manipulating models. The conventional approach to mathematical modeling reflects an implicit quest for Newtonian-type laws embedded in models with stable parameters which at least in the social sciences are never found (Boulding 1992, p. 57). Instead, the Expansion Methods philosophy argues that models should be related to substantively relevant contexts, and that their potential drift across these contexts should be investigated, tested for, and theorized. The central hypotheses of this paper are (i) that the relationship between exports and growth is different at different development levels (namely, it drifts with respect to development); and (ii) that the relationship between rates and levels of development drifts with respect to export growth. Let us apply the Expansion Method to develop a “terminal” model suited for testing simultaneously these hypotheses. Denote by y the percentage rate of growth over time of GDP per capita, Y, and by x the percentage rate of growth over time of exports, X. Namely, y = (l/Y)dY/dt, and x = (l/X)dX/dt.

Jessie Poon

/ 43

Define the following primal initial model relating economic growth, y, to export growth, x: y = a,

+ a,x + E

(1)

where E ( E ) = 0 and vaT(E) = u2.Redefine the parameter a, and a, into quadratic functions of level of economic development on the basis of the following primal expansion equations:

a,

=

coo

a, = c,,

+ COILY+ cozLY2

(2)

+ C,,LY + Cl2LY2

(3)

where LY is the logarithm of GDP per capita. Resubstituting equations (2) and (3) into (1) produces the following terminal model:

+ COILY+ C,,LY2 + c10x + c,,x.LY + C12X.LY'2+

y = coo

E .

(4)

This terminal model is also arrived at from a dual initial model expressing economic growth as a quadratic function of economic development: y = b,

+ b,LY + b,LY2 +

E .

(5)

To this effect, expand the parameters b,, b,, and b2 of (5) into linear functions of export growth rates by the dual expansion equations (6), (71, and (8).

b,

=

b,

= c,,

b,

coo

+ c,,x;

+ c11x; = co2 + c 1 2 x .

(6)

(7) (8)

If the expanded b,, b,, and b, are resubstituted back into (5), the terminal model (4) is obtained again. The expansion equations (2) and (6) define intercept expansions since they expand the intercept terms of the primal and dual initial models. The equations (3), (7), and (8) define slope expansions, namely, expansions of the slope coefficients in the primal and dual initial models. Terminal models can be produced by expanding only the intercepts, only the slope of the primal initial model, only the slopes of the dual initial model, or both intercepts and slopes. The expansion of both intercepts and slopes generates terminal model (4), from which the other terminal models can be produced by restricting some parameters to zero. In the empirical analyses that follow, the terminal models produced by expansions of the intercepts, of the slope in the primal initial model, of the slopes in the dual initial model, and of both intercepts and slopes will be estimated and contrasted. Hypothesis tests on the parameters of the terminal models estimated can be used to determine whether the primal and/or dual initial models hold and drift. The primal/dual hypotheses investigated here can be summarized as follows: the primal relationship between economic growth and export growth is hypothesized to drift with development levels; the dual relationship between economic growth and development levels is hypothesized to drift with export growth.

44

/ Geographical Analysis

TABLE 3 Regression Results of the Primal-Dual Initial Equations Initial Equation

Primal

Independent Variable

Parameter Estimate

Intercept

1.009*** 0.131*** 0.21 26.103*** - 21.962* 6.641* - 0.433* 0.10 5.275***

X

R2

Dual

F-value Intercept

LY LY2 R2

F-value

P-Value

0.01 0.00

0.06 0.06 0.09

***, **, *

Significant at 1, 5 and 10 percent F-value is from analysis-of-variance

4. THE ANALYSES

The analyses that follow are based on a sample of forty-nine developing countries extracted from a larger sample for which data were available, by removing two export outliers (Gabon and Jamaica), and the major oil-exporting countries (see Appendix 1). The source data are average annual growth rates of exports and of GDP per capita over the time intervals 1960-70 and 1970-80, and GDP per capita for 1980 in 1980 US dollars. GDP per capita at the midpoints of the two time intervals were obtained by projecting the 1980 data backward. The data for the two time intervals were pooled. The data sources are the Taiwan Statistical Data book (19871, and UNCTAD (1983, Table 6.1; 1990, Tables 1.5, 1.6, and 6.2). The estimates of the primal and dual initial equations are shown in Table 3. They are used as a benchmark to aid the interpretation of the remaining results. Both the coefficients of the primal initial equation (1) are significant at 5 percent. The estimated a, is greater than zero, signifying that the effect of export growth on economic growth is positive. The R-square is 0.21. The estimated a , implies that an export growth of 1 percent is associated with 0.131 of a 1 percent in GDP per capita growth. Expressed differently, a 7.6 percent increase in exports is associated with a 1percent increase in GDP per capita. The intercept term is also significant, indicating that in the absence of export growth, a country still experiences economic growth. The regression coefficients of the dual initial equation (5) are weakly significant at the 10 percent level although the F-value from the analysis-of-variance of 5.275 is highly significant at 1p e r ~ e n tThe . ~ estimated b , is positive and b, is negative confirming that the countries at intermediate development levels tend to grow faster. White’s (1980) tests for heteroskedasticity were carried out for all the regressions reported in this paper. Except for the primal initial equation (1) which gave a marginally positive result at the 5 percent significance level, the rest gave negative results throughout. The t-values for the coefficients of the primal initial model reported in Table 3 are based on White’s heteroskedasticity consistent estimator of the variance covariance matrix. Table 4 shows the estimates of the terminal models obtained by expanding the intercepts and/or slopes of the initial models. Specifically, model (i) is obtained by expanding both intercepts and slopes; model (ii), by expanding the intercept; model (iii), by expanding the slope of the primal initial model; and model (iv), by expanding the slopes of the dual initial model. All of these models suggest the Occurrence of significant drift. Model (i) has the highest adjusted R-square but 3The analysis-of-variancetests the hypothesis that b, = b , = b , = 0.

Jessie Poon

/ 45

TABLE 4 Regressions of Various Terminal Equations lnde endent variab

(i)

Parameter Value

Expansions of intercept and slope Intercept LY LY2

13.740 - 4.601

0.403

- 2.243

X

x.LY x.LY2 Adjusted R 2 (ii) Expansion of primal model's intercept Intercept LY LY 2

0.720*

- 0.054*

0.50 0.46 0.39 0.13 0.10 0.09

0.238 - 16.673

X

Adjusted R 2 F-valuea (iii) Expansion of primal model's slope parameter Intercept X

P-Value

4.913 - 0.330 0.115*** 0.227 2.388* 1.153***

- 1.762**

x.LY 0.531** x.LY2 -0.037** Adjusted R 2 0.231 F-valuea 7.505*** (io) Expansion of dual model's slope parameters Intercept - 11.107 LY 2.842 LY2 - 0.146 x.LY 0.065* x.LY2 - 0.007 Adjusted R 2 0.227 2.008* F-Value"

0.13 0.13 0.17 0.00

0.00 0.04 0.03 0.04

0.35 0.45 0.62 0.07 0.17

NOTES: 'Reduced model F-test (Johnston 1984)

***, **, *

Significant at 1.5 and 10 percent

most of its coefficients are not significant. The model with the highest R-square and in which all the coefficients are significant is model (iii), which was selected as the result of the analysis. In terms of the primal estimated expansion equations it involves, model (iii) can be rewritten as follows: y = 6,

+ 6,x + e ;

6,

=

1.153;

8,

=

-1.762

+ 0.531LY - 0.037LY2

Model (iii) can be rewritten also in terms of its estimated dual expansion equations: y =

i0=

6, + S,LY + &,LY, + e ; 1.153 - 1 . 7 6 2 ~ ;

A

b,

=

0.531~;

=

-0.037~.

A

b,

46

/ Geographical Analysis ixport Growth Parameter

0.16

(A)

0.14

0.12

0.1

0.08

0.06

0.04

0.02

A

* 1

0 0

Expanded

i

Unexpanded

I

I

I

1

I

1

2

3

4

5

6

GDP Per Capita (’000s$1 FIG. 1. Change in Effect of 1 Percent of Export Growth on Economic Growth with Economic Development Levels

In the sections that follow I will focus on the significance and meaning of the primal and dual drift uncovered by this analysis. Of the two parameters of the primal initial model, d o does not drift with development levels, while a^, does. Let us recall that a, specifies the effect of 1 percent increase in export growth on GDP per capita growth. My results indicate that this effect is a function of development levels. Figure 1 shows graphically the drift of a^, superimposed on the horizontal base line representing the a^, coefficient in the unexpanded initial model. For the sample under study, a^, is a parabola with a maximum at $1,350. Specifically, the figure shows that the effect of export growth on economic growth is increasingly positive from lower to intermediate development levels and then levels off and declines at high development levels. The results support the notion that some minimum level of development is required before export-led growth can materialize. Also, the sharp rate of increase in the estimated a, before peaking points to the great advantages of export growth during an intermediate economic stage. For exports to be growth-inducing, the stage of a country’s economic development does indeed matter. The increasing values of a^, during the lower to intermediate stages of a country’s development may reflect the accumulation of industrial experience in a country’s transition to modernization. As Kavoussi (1985) has observed, producers face difficulties in expanding foreign sales during the initial phase of export-promotion industrialization because of inexperience in meeting quality and delivery requirements of their foreign clients, as well as in promoting products among consumers whose tastes are different from their domestic clientele. Over time, the necessary production and marketing skills for foreign markets accumulate and export expansion becomes easier.

Jessie Poon

/ 47

TABLE 5 Summary Statistics on

81

i1 Mean

Standard Deviation

Minimum

Maximum

0.114

0.044

- 0.049

0.150

The decline of a^, after peaking at $1,350 indicates a trailing off in the impact of export growth on economic growth as a country moves to higher income levels. Two explanations for this finding may be offered: (i) Continued factor productivity is necessary in order to sustain economic growth. This requires a successful transformation of the export commodity structure from primary to manufactured exports. Kavoussi (1984) observes that at higher income levels, the positive effects on productivity of export expansion in primary commodities tend to vanish. Exports of manufactures, however, have been dominated by only a few developing c o u n t r i e ~(ii) . ~ A country tends to experience shifting comparative advantages over time. The emergence of the younger or second-tier newly industrializing countries (NICs) over the older first-tier NICs illustrates this. Second-tier NICs are defined by their ability to engage in exports that were previously dominated by the older NICs (Athukorala 1989).’ For the older NICs, the gains from reallocation due to exports are likely to be eliminated in the long run and Pack (1988) emphasizes the need to increase technical efficiency in the manufacturing sector. Table 5 provides a summary of the variation in a,. Its mean is positive at 0.114 illustrating that there is no change in the average positive pattern observed by other authors. However, the individual values of 6, for the sample range widely from -0.049 to 0.15 reflecting the factor of income levels. The drift of the dual initial model will be considered next. The unexpanded estimated initial model represents the average trajectory of developing countries’ growth in relation to their development levels in the absence of specific effects from export growth. The estimat;lon of the du%lexpansion equations clearly show that the dual initial model y = b, b,LY b,LY2 does drift with respect to x, namely, it holds with parameter values that are a function of export growth. Three portraits capturing the effects of low, middle, and high export growth on the original economic growth-development relationship are presented here. Low, intermediate, and high export growth values were determined by averaging the export growth for countries assigned to these three categories. In Figure 2 the unexpanded estimated dual initial model is portrayed by the “base curve,” and the estimated dual initial models produced by low, intermediate, and high export growth are superimposed on it. All the curves in Figure 2 show that the rate of economic development tends to be higher for countries at an intermediate development levels and tends to be lower at low and high development levels. However, the curves also show that the rate of growth of exports makes a considerable difference in these trajectories. Higher export growth is clearly associated with the attainment of higher maximum rates of development, with a much faster decline of these rates as higher develop-

+

+

4Balassa (1978b) notes that only a dozen newly industrializing countries accounted for 68 percent of the exports of manufactures from developing countries in 1973. ’The observation that countries such as Brazil, Greece, Mexico, Portugal, Singapore, South Korea, Spain, Taiwan, and Yugoslavia had become prominent exporters of manufacturing goods in the 1960s has led to the term “newly industrializing countries” (NICs). A second grou of developing countries followed suit along the same line in the 1970s occu ying the positions left gy the older NICs as the latter’s comparative advantages shifted. This led to t f e terms “first” and “second’ tier NICs (OECD 1982).

48

/ Geographical Analysis 6

GDP Per Capita Growth (%b)

I

A

Low Export

I

0

1

2

3

4

5

6

GDP Per Capita (‘000s $1

FIG. 2. The Effect of Low, Middle, and High Export Growth on the Relationship between Rates and Levels of Development

ment levels are attained, and with lower rates of economic growth at lower development levels. Let us contrast the base curve to the other curves in the figure. The base curve is produced by the estimate of the unexpanded dual initial model, and depicts the average growth-levels relationship for development when the effects of exports growth are disregarded. Clearly, the base curve underestimates substantially the economic growth that can be attained in the presence of high rates of export growth, and at the same time suggests an unwarranted optimism at the rates of development that can be expected when the growth of exports is intermediate to low. However, this curve also overestimates the. economic growth at very low development levels. The implication that may be drawn from the above is that lower to moderate export growth appears to be more desirable when a country is at a low development stage. For example, at a GDP per capita of $150 the economic growth variable, y , is 1.03 percent when export growth is low, but only 0.57 percent when export growth is high. Clearly, the Marxist notion that exports may translate into a rape of the exporting country’s resources comes to mind in this connection, especially so, since it is at very low levels of development that export growth, in terms of these results, is perverse. Similarly, at high levels of economic development, the beneficial effects of high export growth are not as apparent. One possible explanation for this is that the contribution of exports becomes relatively less important as a country’s economy increases in size. Kuznets’ work (19641, for example, has indicated lower foreign trade proportions for larger countries, size here being defined in terms of gross national product. A decline in the importance of the export sector with increased size is also implied by Erickson (1989) who has observed a small foreign export effect on employment in a large domestic economy like the United States. At any rate, the effect of high export growth on the economic growth-development relationship is not as conspicuous at low or high income levels. Figure 2

Jessie Poon

/ 49

indicates that the effect of high export growth on the economic growth-development relationship is still the most desirable of the three export curves at comparable income levels beyond $180 (where the three export curves intersect) for the range of incomes graphed. This confirms again the merit of export-promotion development policies only above minimum income levels. In a capsule, the analysis reported here shows that the primal exports-growth relationship drifts with the levels of economic development, and that the dual relationship between rates and levels of economic development drifts with export growth. 5. CONCLUSION

The export-promotion development strategy has been touted as a universally valid prescription for rapid economic growth. The empirical support for export-promotion was sought and found in econometric analyses of relationships between economic growth and export variables. The presupposition that exports and especially export growth are positively related to economic growth has somehow emerged from these analyses, despite of the wide variety of results that they actually produced. In fact, a closer scrutiny suggests that empirical support for the notion that the export-promotion strategy is universally valid does not exist. Following the conventional econometric practice, the exports-growth relationships have been estimated under the implicit presupposition of parametric invariance. This presupposition is wrong, and it leads to conclusions unwarranted in their generality. The model-context perspective advocated by the Expansion Methods research philosophy and applied in this paper is the appropriate one to follow in the investigation of exports-growth relationships. This perspective leads to asking the following question: across which contexts can the export growth relationship be presumed to drift, namely, to hold with different parameter values? The context investigated here is development levels. No implication is intended that other contexts could not be usefully investigated. My results indicate that the exports-growth relationship varies with economic development levels, and that, consequently, the relationship between rates and levels of economic development also varies with export growth. These results have substantial implications on the validity of the export-promotion development strategy. The parametric variation of the relationship between economic growth and export growth with respect to development levels implies that exports vary in their effect on a country’s growth according to its stage of development. Specifically, the positive effect of exports on economic growth accelerates markedly as a country moves from a low to an intermediate position on the development continuum. This effect levels off at some point beyond which exports become less and less an engine of growth. Also, the effect does not operate below a low development threshold, possibly because of the relatively high costs of concentrating scarce resources in the export sector at low productive stages. Consequently, the exportpromotion development strategy is a prescription for accelerated economic development only during the intermediate stages of a country’s development cycle. Low to moderate export growth may be more desirable for countries at low development stages. And export-promotion should be at the very least supplemented, if not replaced, by other development strategies in the more-developed developing countries. I would like to pinpoint what is gained by switching from the conventional presupposition of invariance to the Expansion Methods model-context perspective in the investigation of exports-growth relationship. Suppose that this relationship is

50

/ Geographical Analysis

estimated under the presupposition that if it holds, it will hold similarly under all circumstances. Then, we may be led to conclude also that the export-promotion development strategy is a universal prescription for faster economic growth. If instead the variation of the relationships across a context is established, the points in the context for which the export-promotion development strategy is not warranted are also established. In conclusion, the study reported here suggests that the export-promotion development strategy is especially important for countries at intermediate development levels. Complementary development strategies warrant serious consideration at development stages in which the export sector-lacks the requisite growth-propulsive qualities. APPENDIX I Export Growth, GDP per Capita Growth, and CDP per Capita Export Growth (%)'

Country

Ar entina Bofivia Brazil Chile Colombia Costa Rica Cyprus Dominican Rep Ecuador EWvador Gabon Ghana Greece Guatemala Haiti Honduras Hong Kong Indonesia Israel Ivory coast Jordan Liberia Madagascar Malawi Malaysia Mauritius Mexico Morocco Panama Pakistan Paraguay Peru Philippines Portugal Senegal Singapore South Korea Spain Sri Lanka Taiwan' Thailand Tunisia Tanzania Turkey Uruguay Venezuela Zaire Zambia Zimbabwe

GDP Per Ca ita Growth t b '

GDP Per Capita

($Y

1960-70

1970-80

1960-70

1970-80

1980

4.8 15.2 7.2 10.2 4.1 10.6 7.9 1.3 3.8 4.5 7.0 11.3 2.2 12.2 10.4 0.6 12.2 14.5 1.7 13.6 11.7 13.1 13.1 5.5 9.9 4.3 2.1 6.0 3.7 14.9 10.3 7.2 8.5 7.5 12.1 3.1 3.3 39.6 13.3 - 1.4 23.2 5.9 4.7 5.0 6.0 3.6 1.2 6.1 13.4 5.7

18.0 19.9 21.7 16.1 19.6 18.3 17.9 15.5 30.4 12.9 19.1 34.1 12.1 23.2 19.6 19.6 17.9 22.4 35.9 21.6 23.0 34.9 10.8 12.3 16.4 24.9 21.0 25.7 16.1 13.0 13.3 18.8 14.8 17.5 14.2 15.4 28.2 37.2 23.7 13.6 28.6 24.7 27.0 7.5 16.2 17.8 20.3 6.2 3.5 12.9

2.7 3.3 2.4 2.2 2.1 2.9 5.8 1.6 1.8 2.0 2.5 5.5 -0.1 6.9 2.7 - 1.3 2.0 11.0 1.1 4.5 3.8 3.7 2.5 1.1 2.2 3.0 - 0.7 3.9 1.1 4.7 3.3 1.9 1.9 2.1 6.1 - 2.0 6.8 6.3 5.9 2.5 6.8 5.1 2.2 4.7 3.4 0.2 2.2 0.5 5.2 0.7

0.7 2.0 5.8 - 0.2 3.1 2.9 1.8 3.9 5.9 4.8 1.2 7.8 - 3.0 3.6 2.9 2.5 1.3 5.2 5.4 1.5 2.6 4.4 - 1.0 - 2.0 3.0 5.3 5.8 3.2 2.8 1.9 2.1 5.1 0.5 3.7 3.6 - 1.3 6.8 7.2 2.4 3.0 7.3 4.6 4.1 1.1 3.0 2.8 1.3 -3.1 - 1.9 - 1.7

5667 988 1967 2528 1245 2190 3230 1106 1417 464 706 6359 353 4302 1081 244 674 4503 460 5321 1315 911 507 366 248 1724 1002 2662 876 1788 339 1450 1082 696 2493 513 4525 1534 5625 281 2325 702 1372 275 1254 3512 4024 113 671 744

Sources: tUNCTAD, 1990, Tables 1.5, 1.6, and 6.2. UNCTAD, 19tX3,Table 6.1. 'Taiwan Statistical Databook, 1987.

Jessie Poon

/ 51

LITERATURE CITED Adelman, Irma (1984). “Beyond Export-led Growth.” World Development 12, 937-49. Adelman, Irma, and Sherman Robinson (1989). “Income Distribution and Development.” In Handbook of Development Economics, vol. 2, pp. 950-1003. New York: North-Holland. Athukorala, Premachandra (1989). “Export Performance of ‘New Exporting Countries’: How Valid Is the Optimism.” Development and Change 20, 89- 120. Balassa, Bela (1978a). “Exports and Economic Growth: Further Evidence.” Journal of Development Economics 5, 181-89. (1978b). “Export Incentives and Export Performance in Developing Countries: A Comparative Analysis.” Weltwirtschafliches Archiv 114, 24-60. (1985). “Exports, Policy Choices, and Economic Growth in Developing Countries after the 1973 Oil Shock.” Journal of Development Economics 18, 23-35. (1988). “The Lessons of the Asian Development: An Overview.” Economic Development and Cultural Change 36 (supplement 3). (1989). “Outward Orientation.” In Handbook of Development Economics, vol. 2, edited by H. B. Chenery and T. N. Srinivasan, pp. 1645-90. Amsterdam: North-Holland. Banerji, R., and James Riedel (1980). “Industrial Employment Expansion under Alternative Trade Strategies: Case of India and Taiwan: 1950-1970.” Journal of Development Economics 7, 567-77. Bhagwati, Jagdish (1978). Foreign Trade Regimes and Economic Development: Anatomy and Consequences of Exchange Control Regimes. Cambridge, Mass.: Ballinger. (1988). “Export-Promoting Trade Strategy: Issues and Evidence.” The World Bank Research Observer 3, 27-57. Boulding, Kenneth E. (1992). “Appropriate Methodologies for the Study of the Economy.” In Praxiologies and the Philoso hy of Economics: The International Annual of Practical Philosophy and Methodology, vol. 1, e d i t e l by J. Lee Auspitz et al., pp. 43-59. News Brunswick: Transaction Publishers. Bruton, Henry (1989). “Import-Substitution.” In Handbook of Development Economics, vol. 2, edited by H. B. Chenery and T. N. Srinivasan, pp. 1601-44. Amsterdam: North Holland. Casetti, Emilio (1972). “Generating Models by the Expansion Method: Application to Geographical Research.” Geographical Analysis 4, 81-91. (1982). “Mathematical Modelling and the Expansion Method.” Statistics or Geographers and Social Scientists, edited by R. B. Mandel, pp. 81-95. New Delhi: Concept Pub ishing. (1986). “The Dual Expansion Method: An Application for Evaluating the Effects of Population Growth on Development.” IEEC. Transaction on Systems, Man, and Cybernetics SMC-16. Casetti, Emilio, and John Paul Jones 111 (1992). “An Introduction to the Expansion Method and to Its Applications.” In Applications of the Expansion Method, edited by John Paul Jones 111 and Emilio Casetti, pp. 1-9. New York: Routledge. Cline, William (1982). “Can the East Asian Model of Development Be Generalized?” World Development 10, 81-90. Emery, Robert F. (1967). “The Relation of Exports and Economic Growth.” Kyklos 20, 470-85. Erickson, R. (1989). “Export Performance and State Industrial Growth.” Economic Geography 65, 280-92. Feder, Gershon (1982). “On Exports and Economic Growth.” Journal of Development Economics 12, 59-73. Fosu, Augustine Kwasi (1990). “Exports and Economic Growth: The African Case.” World Development 180, 831-35. Gerschenkron, Alexander (1962). Economic Backwardness in Historical Perspective. New York: F. A. Praeger. Helleiner, G. K. (1986). “Outward Orientation, Import Instability, and African Economic Growth: An Empirical Investigation.” Theory and Reality in Development: Essays in Honor of Paul Streeten, edited by Sanjaya Lall and Frances Stewart, pp. 139-53. London: Macmillan. James, William E., Seija Naya, and Gerald M. Meier, eds. (1989). Asian Development, Economic Success and Policy Lessons. Madison: University of Wisconsin. Johnston, R. (1984). Econometric Methods, 3d ed. New York: McGraw-Hill. Joint ECLA/UNIDO Industrial Development Division (1986). “Thoughts on Industrialization, Linkage, and Growth.” CEPAL Review (28), 49-66. Kavoussi, Rostam (1984). “Export Expansion and Economic Growth: Further Empirical Evidence.” Journal of Development Economics 14, 241-50. (1985). “International Trade and Economic Development: The Recent Experience of Developing Countries.” Journal of Developing Countries 19, 379-92. Krueger, Anne (1978). Foreign Trade Regimes and Economic Development, Liberalization Attempts and Consequences. Cambridge, Mass.: Ballinger. (1985). “The Experiences and Lessons of Asia’s Superexporters.” In Export-Oriented Deuelopment Strategies: The Success of Five Newly Industrializing Countries, edited by Vittorio Corbo, Anne Krueger, and Fernando Ossa. Boulder: Westview Press.

f

52

/ Geographical Analysis

(1988). “The Relationshi between Trade, Employment, and Development.” In The State of Development Economics, editel by Gustav Ranis and T. Paul Schultz, pp. 357-85. New York: Basil Blackwell. Kuznets, Simon (1964). “Quantitative Aspects of the Economic Growth of Nations: Level and Structure of Foreign Trade.” Economic Development and Cultural Change 13, 1-105. Lewis, W. Arthur (1969). Aspects of Tropical Trade, 1883-1967. Uppsala: Wiksell. (1979). The Evolution of the International Economic Order. Princeton: Princeton University Press. -(1980). “The Slowing Down of the Engine of Growth.” American Economic Review 70, 555-64. (1988). “Reflections on Development.” In The State of Develo ment Economics, edited by Gustav Ranis and T. Paul Schulz, pp. 13-30. New York: Basil Blackwet. Little, Ian, T. Scitovsky, and M. Scott (1970). Industry and Trade in Some Developing Countries: A Comparative Study. New York: Oxford University Press. Maizels, Alfred (1968). Exports and Economic Growth of Developing Countries, London: Cambridge University Press. Michaely, Michael (1977). “Exports and Economic Growth: Comment.” Kyklos 4, 49-53. Michalopoulos, Constantine, and Keith Jay (1973). “Growth of Exports and Income in the Developing World: A Neoclassical View.” AID Discussion Paper No. 28, Washington, D.C.: Agency for International Development. Morawetz, David (1977). Twenty-five Years of Economic Development, 1950 to 1975. Washington, D.C.: The World Bank. Moschos, Demetrios (1989). “Export Expansion, Growth, and Level of Economic Development.” Journal of Development Economics 30, 93-102. North, Douglas C. (1961). The Economic Growth of the United States, 1790-1860. Englewood Cliffs, N.J.: Prentice-Hall. Nurkse, Ragnar (1961). Patterns of Trade and Development. New York: Oxford University Press. OECD (1982). “Developing Country Exports of Manufactured Products: The Experience of the Second-Tier‘ Countries.’’ Paris, Development Cooperation, 23-132. Pack, Howard (1988). “Industrialization and Trade.” In Handbook of Deuelopment Economics, vol. 1, edited by H. B. Chenery and T. N. Srinivasan, pp. 333-80. Amsterdam: North-Holland. Prebisch, Raul (1962). “The Economic Development of Latin America and Its Principal Problems.” Economic Bulletin for Latin America 7 (1). (First published in 1949). (1984). “Five Stages in My Thinking on Development.” In Pioneers in Development, edited by Gerald M. Meier and Dudley Seers, pp. 175-91. New York: Oxford University Press (published for The World Bank). Ram, Rati (1987). “Exports and Economic Growth in Developing Countries: Evidence from Time-Series Data.” Economic Development and Cultural Change 36, 51-72. Rana, Pradumna B. (1985). E x p w t s and Economic Growth in the Asian Region. Economic Staff Paper Number 25. Manila, Phillipines: Asian Development Bank. (1988). “Exports, Policy Changes and Economic Growth in Developing Countries after the 1973 Oil Shock.” Journal of Development Economics 28, 261-64. Ranis, Gustav (1985). “Can the East Asian Model of Development Be Generalized: A Comment.” World Development 13, 543-45. Reynolds, Lloyd G. (1983). “The Spread of Economic Growth to the Third World.” Journal of Economic Literature 11, 941-80. Riedel, James (1987). Myths and Reality of External Constraints on Development. BroolGeld, Vt.: Gower. Russett, B. Alker, Jr., K. W. Deutsch, and H. D. Lasswell, (1964). World Handbook of Political and Social Indicator. New Haven, Conn.: Yale University Press. Salvatore, Dominick, and Thomas Hatcher (1991). “Inward-oriented and Outward-oriented Trade Strategies.” Journal of Development Studies 27, 7-25. Singer, Hans (1950). “Distribution of Gains between Investing and Borrowing Countries.’’ American Economic Review, Papers and Proceedings 40, 473-85. (1984). “Terms of Trade Controversy and the Evolution of Soft Financing: Early Years in the U.N.” Pioneers in Development, pp. 227-303. New York: Oxford University Press (published for The World Bank). (1991). “Terms of Trade, New Wine and New Bottles?’ Development Policy Review 9,339-51. Singer, Hans, and Patricia Gray (1988). “Some New Data, Trade Policy, and Growth of Developing Countries.” World Development 16, 395-403. Spetter, Henry (1970). Inward Looking and Export-Oriented Industrialization in Developing Countries. Budapest: Center for Afro-Asian Research of the Hungarian Academy of Sciences. Syron, Richard F., and Brendan M. Walsh (1968). “The Relations of Exports and Economic Growth; A Note.” Kyklos 21, 541-45.

Jessie Poon

/ 53

Syrquin, Moshe (1986). “Productivity Growth and Factor Reallocation.” In Industrialization and Growth, edited by H. B. Chenery, S. Robinson, and M. Syrquin. New York: Oxford. (1988). “Patterns of Structural Change.” In Handbook of Development Economics, vol. 1, edited by H. B. Chenery and T. N. Srinivasan, pp. 203-74. Amsterdam: North-Holland. T a d e , E. J., and Emilio Casetti (1990). “The Model-Context Problem and the Expansion Method.” Paper presented at the Annual Meeting of the Association of American Geographers. Toronto: Canada. Taiwan Statistical Data Book (1987). Republic of China: Council for Economic Planning and Development. Tyler, William G. (1981). “Growth and Export Expansion in Developing Countries: Some Empirical Evidence.” Journal of Development Economics 9, 121-30. UNCTAD (1983). Handbook of International Trade and Development Statistics. New York. -(1990). Handbook of International Trade and Development Statistics. New York. Westphal, L. (1978). “The Republic of Korea’s Experience with Export-Led Industrial Development.” World Development 6, 347-82. White, H. (1980). “A,Heteroskedasticity-ConsistentCovariance Matrix Estimator and a Direct Test of Heteroskedasticity. Econometrics 48, 817-38. World Bank (1987). World Development Report, 1988. Washington, D.C.