This final chapter of Part Three is aimed at providing a more general test of the theoretical framework proposed in Chapter 7. However, since the model proposed there was very stylized, the procedure of testing it necessarily involves some 'creativity' by transforming the general consequences of the model into testable hypotheses. In order to do so, two different strategies will be used. First, some equations will be derived on the basis of the model, and these will be estimated using the data set and some of the results from the previous chapter. This provides a general approach to the subject, which pays no attention to results for specific countries. This is were the stylized facts from Chapter 4 will be brought back into the analysis: Among other things, the idea of structural differences will be applied to the case of uneven growth.

The second approach is more case study-oriented, and it concentrates on the Asian countries in the sample, which were shown to be 'prime' examples of catching up in Chapter 4. The current analysis tries to give an in-depth overview of the observed performance of these countries in light of the preceding analysis. However, the reader should keep in mind that it is not the aim of this part to give a complete overview of development in Asian NICs. Instead, the analysis will focus on isolated parts which are directly relevant to the preceding chapters.

9.1. A General Test of the Relation Between Competitiveness, Structure and Growth Rate Differentials

In applying the model from Chapter 7 to actual data, a number of problems arise. The first problem is concerned with the concepts used in the model. The main force determining the growth rate of a country was the balance of payments, which was assumed to be in equilibrium at all times. During the presentation of the model it has already been admitted that this assumption does not comply with the real-world facts. What one observes are balance of payments deficits and surpluses, and as a result, accumulation of debt in some countries. Nevertheless, Thirlwall (1979) and Fagerberg (1988a) have shown that the balance of payments restriction to growth rates does make sense in an empirical setting. Therefore, although balance of payments equilibrium is seldom achieved, growth rates seem to converge to the value that is consistent with external equilibrium, at least for the countries investigated by Thirlwall and Fagerberg.

However, both Thirlwall and Fagerberg used aggregate data to test their models. As was shown in Chapter 7, the country-wise differences in elasticities observed by them can (at least theoretically) be explained by the production and consumption structure of the domestic economy. This is the main reason why the analysis in Chapters 7 and 8 was extended to include the sectoral level. However, this poses another problem: The data used in the previous chapter are available for the manufacturing sector only, and, as a result, many sectors of the economy, such as agriculture, mining, building, transport, and other forms of services, were ignored. And even though large parts of the services sector are nontradeable, concentrating on manufacturing alone leads to ignoring large parts of the trade balance. Therefore, even if the balance of payments restriction to economic growth is relevant, it cannot be used in the narrow manufacturing perspective adopted here.

Thus, it is imperative to develop another way of exploring the empirical consequences of the model proposed in Chapter 7. To do this, assume for the moment that there is only one market for each of the products of the industrial sectors identified in the previous chapter. In other words, the relevant market for each producer is the total world market, irrespective of its location. This means that the detailed specification of the relation between competitiveness and growth through the import and export sides of the trade balance will be 'skipped'. Then, using a simple definition, each country's rate of growth of production in a specific sector over a specific period will be the sum of the growth rate of its market share and the growth rate of volume of the market.

In this equation, the symbols are as defined in previous chapters, and M is the size of the market. As before, it is assumed that the movement of the market share is determined by competitiveness, so that a country whose competitiveness is exactly equal to the average will grow as fast as the market volume. This country is denoted by *.

The next step is to write an expression for the aggregate growth rate of countries. Obviously, the aggregate growth rate can be found by adding together the sector-wise growth rates, taking sector shares as weights. In mathematical form, this is written as follows.

6 =V с 0 =Y* cj i +V о M (IX.3)

i i 1

Obviously, for the 'average competitive' country (*), the aggregate growth rate reduces to the following.

Combining the last two equations, the growth rate differential between an individual country and the average country growth rate can be written as follows.

(2,-0 '-E ^aA⁺E < v°P4

197

This equation shows that the growth rate differential is the sum of two partial effects, the first can be attributed to competitiveness, and the second to the structure of production. Using the evolutionary equation used in the two preceding chapters, the last equation can be rewritten as follows.

,-0 '=£

00 '=£ <°Д: <VV VD)⁺E Ч°?Ч «x

The first term on the rhs of equation (IX.6) is due to competitiveness, while the
second one is due to the production structure. ...

However, the irrealistic assumption of one world market leads to the necessity of carefully interpreting this equation. First, the competitiveness part does not take into account all sorts of factors that might prohibit a successful transformation of high competitiveness into high growth. The most obvious of these factors is the existence of trade barriers in the form of protectionist measures, and the space dimension, which leads to trade flows from and to one specific country that are unequally distributed over the world. Second, the structural part of the equation models the structural problem from the supply side, while it seems to be more logical to model it from the demand side, as in Chapter 7. In other words, in the theoretically preferred approach from Chapter 7, the existence of a structural advantage or disadvantage was determined by the country's and the world's consumption structures, while here it are the production structures that matter. Obviously, at the total world level, the consumption and production structures must be equal, but as the simulation results in Chapter 7 showed, for separate countries that are trading with other countries, the two are likely to differ, due to specialization. Therefore, part of the equation is misspecified with regard to the impact of competitiveness on growth through the import (i.e., demand) side of the economy.

Nevertheless, equation (IX.6) can be tested using the data and results for ф estimated in the previous chapter. Before doing so, a regression relating growth rate differentials of manufacturing production and total GDP (denoted by Y) is carried out. This equation gives an idea as to what extent the results for manufacturing have significance for the economy as a whole. The equation used is as follows¹.

y.-Y'=_Va_o(0,-Q')

The outcome of this regression is the following.

a₀ = 0.437 (22.02*") Y₀ = 0.001 (0.88)

¹ The data set used in the regressions in this section is partly the same as that used in the previous chapter. Manufacturing variables not used there (such as the growth rate of output) are taken from the same source (UNIDO Industrial Statistics Database). Data for GDP in this chapter are taken from Summers and Heston (1991). Countries used and the period involved are also the same (1963-1987). Starred values refer to (weighted) sample averages.

198

n = 652

adj. R² = 0.43

A Chow-test for structural change between the periods before and after 1973 is not significant. These results show that there is a strong correlation between the growth rate differential with respect to total output and the growth rate differential of manufacturing output. Of course, this correlation is not surprising, since manufacturing output is a part of total output. Nevertheless, the value of the estimated coefficient, which is clearly smaller than one, shows that there is a general tendency in the data towards specialization between the manufacturing and nonmanufactuhng countries. The reason for this is that the estimated value of a1 shows that large positive values of growth rate differentials in manufacturing tend to go hand in hand with below-average growth (i.e., negative growth rate differentials) in nonmanufacturing sectors of GDP. Large negative values of the growth rate differential in manufacturing tend to go together with above-average growth in nonmanufacturing sectors. As a result, the growth rate differential induced by the manufacturing sector is partly offset by the other sectors. In other words, the manufacturing sector has induced a tendency towards divergence, while this has been partly offset by a converging tendency in nonmanufacturing sectors.

Now that the relation between manufacturing and total growth rate differentials is clear, attention can be shifted towards explaining the growth rate differentials. In order to do so, define the following variables.

СХЦХ.+М^/У. (IX.10)

0 stands for competitiveness, and is defined according to equations (IX.5) - (IX.6). The фs are taken from the empirical analysis in the previous chapter². S is the effect of the production structure, and is also defined as in the equations above. О measures the openness of the economy. This variable is used to correct the relations above for different degrees of openness. Several specifications will be used to do this, which will be discussed below in more detail. H is a measure of the degree of specialization of the economy. It is defined as the variance of the sectoral shares in manufacturing output around the sample means, so that high

² фs used are moving averages of those obtained in the estimations for Table VIII.ll, so that competitiveness consists of the wage rate, labour productivity and a scale factor. For the exact values of the фs used, the reader should refer to the floppy disk. The program that should be started is ELA5.EXE, and variable names are as follows: Ф^е, = CNj; ф₁,_Ьццгp^^miy,, ~ CPj; Фч-арпм,! = CWj (j refers to a sequence number in Appendix IV.2).

199

values correspond to a high degree of specialization. This variable is meant to pick up the effects of increasing returns to scale due to specialization, which were explained in the discussion of the simulation results of the model in Chapter 7. In principle, these scale effects are assumed to be included in the measures of competitiveness used, but in order to see if any of these effects are not captured through 0, the H variable is included in the regressions. Although the definitions of Я and S are alike, the correlation between the two is low (-0.03). This illustrates that H and S measure two different things: H is related to the static structure of the economy, which is assumed to provide opportunities for dynamic scale effects while S measures the dynamic structural advantages related to market demand. к stands for the investment intensity, and is included to take into account some aspects of competitiveness that are not included in the definition of 0.

These variables will be used in a number of different equations. Since the basic equation derived above (IX.6) assumes that all economies are completely open, some additional specifications will be tested that relax this assumption. The basic idea behind these other forms is that the relation between competitiveness and growth rate differentials is stronger for economies that are more open. Thus, one can assume that in a regression of the type (IX.6), the estimated coefficient for competitiveness varies with openness. One way of taking this into account is by estimating an equation with competitiveness multiplied by openness (O-0) as one of the independent variables. Another possibility is to specify a nonlinear (in the parameters) relation between openness, competitiveness and growth. Both approaches will be followed below.

First, some linear (in the parameters) equations will be estimated. The estimates will be done both for equations explaining the growth rate differential in manufacturing, and GDP. The results for the linear equations are in Tables IX.l and IX.2. The tables show that the overall explanatory power of the regressions is rather weak. However, the coefficients are (highly) significant in most cases, which indicates that although the variance explained is low, there is a significant relationship of the kind assumed. Chow-tests for structural change between periods before and after 1973 are not significant.

Turning to the individual equations, the following can be said. In Table IX.l, equation (i) is the purest form of the hypothesis derived above (IX.6). The parameter estimate of 0 is smaller than one, indicating that there is indeed a factor which prohibits the differences in competitiveness to be transformed into differences in growth rates completely. This is probably a mixed effect of the competitiveness measures being less than perfect and of the omission of the openness effect. The estimated parameter of the structural term S is larger than one, which is hard to explain from the point of view of the above equations.

Equation (ii) tries to correct for the openness by assuming that the slope of the basic equation (as in i) varies with openness. The minimum value of О observed (around 4%) corresponds to an estimated coefficient of around 0.015, while the maximum О value (around 245%) yields a slope of around one. Thus, these results indicate that the slope of the competitiveness variable varies between zero and one, with the extremes of this interval reached for the most closed and open

200

economies in the sample.

Equations (iii) and (iv) are basically the same, but introduce the specialization term. It is shown that higher specialization leads to higher growth. The coefficient of 0O becomes smaller and insignificant, while the (adjusted) R² in (iv) gets smaller by including the extra variable. This effect might be partially due to multi-collinearity (the correlation between 00 and H is close to 0.4). In any case, this shows that part of the effect of the openness variable in (ii) is due to specialization increasing with openness. The constant in (iii) - (iv) becomes smaller, while the other coefficients remain more or less the same.

Table IX.1. Estimation results for linear equations explaining growth rate differentials for manufacturing output (n=652)

No	С	CO	S	H	К	с	R'
i	0.27 (2.94 "•)		1.90 (5.60 "*)			0.016 (7.27 ***)	0.06
ii		0.42 (3.09 "•)	1.87 (5.52 ***)			0.016 (7.30 ***)	0.06
iii	0.24 (2.60 "*)		1.96 (5.89 ***)	0.33 (5.02 •••)		0.008 (2.76 ***)	0.09
iv		0.19 (1.33)	1.95 (5.83 ***)	0.31 (4.37 ***)		0.009 (3.28 ***)	0.08
V	0.23 (2.42 •*)		1.86 (5.49 ***)		0.06 (2.01 **)	0.002 (0.24)	0.06
vi		0.38 (2.78 "•)	1.83 (5.41 ***)		(0.06 2.23 •*)	0.0003 (0.04)	0.06
vii	0.17 (1.83 *)		1.91 (5.76 ***)	0.37 (5.49 ***	0.09 (2.98 **)	-0.014 (1.83 *)	0.10
viii		0.10 (0.65)	1.90 (5.73 ***)	0.37 (5.01 ***)	0.10 (3.31 ***)	-0.016 (1.99 **)	0.10

Variants (v) and (vi) show that there is also a significant relation between investment intensity and growth rate differentials, and that the influence of S and 9 is not affected by this. Finally, equations (vii) and (viii) include all the variables, and show the general significance.

Table IX.2 repeats the same relations, but explaining GDP growth rate differentials instead. As could be expected from the estimated relation between GDP and manufacturing growth, the coefficients in these equations are smaller than the ones in (i)-(iv), and so are the R²s. However, the coefficients are also significant, which indicates that the relations are strong enough to survive additional sources of disturbance caused by the relation between GDP and manufacturing output.

201

Table IX.2. Estimation results for linear equations explaining growth rate differentials for GDP (n=652)

No	С	CO	S	H	К	constant	R²
i	0.16 (2.55 ***)		0.52 (2.25 **)			0.007 (4.90 ***)	0.02
ii		0.35 (3.78 ***)	0.49 (2.14 **)			0.007 (4.65 ***)	0.03
iii	0.13 (2.08 **)		0.58 (2.58 ***)	0.31 (6.98 ***)		-0.001 (0.28)	0.08
iv		0.13 (1.36)	0.57 (2.53 ***)	0.30 (6.16 ***)		0.0002 (0.10)	0.08
V	0.13 (1.94 ••)		0.49 (2.12 **)		0.05 (2.53 ***)	-0.005 (0.97)	0.02
vi		0.32 (3.43 ***)	0.46 (2.01 **)		0.05 (2.58 ***)	-0.005 (1.08)	0.04
vii	0.07 (1.13)		0.53 (2.42 **)	0.34 (7.60 ***)	0.08 (3.87 ***)	-0.020 (3.73 ***)	0.10
viii		0.05 (0.54)	0.53 (2.40 **)	0.34 (6.94 ***)	0.08 (4.06 ***)	-0.02 (3.79 ***)	0.10

The influence of openness on the coefficient of 0 can also be estimated by means of nonlinear specifications. Various alternatives were tested, which are all nested in the following equation.

D.=(u+aOf)0.+pS.+Y ^(IX-¹³⁾

Estimating this equation in its least restrictive form (i.e., leaving all the parameters free) does not yield very good results, both in terms of convergence and in terms of f-values of the estimated parameters. Therefore, various special cases of the general equation were estimated, which, in general, produce quite good results.

Table IX.3 lists the results (Chow-tests for structural change are not significant). In equations (i) and (iv), one would expect 1 > u+O⁵ > 0 (close to zero for closed economies, and close to one for open economies), which implies that u<0 and 5>0. Note that this equation assumes that the relation between competitiveness and growth is zero for an economy with a value of O>0. In other words, the point at which competitiveness becomes meaningless lies before the point of a completely closed economy. The performance of the equation for GDP growth is better than for the equation for manufacturing, at least in terms of significance of the coefficients. Although the relation may not be very strong, it is useful to calculate the boundaries of the range for the implied coefficient of 0. These are as follows: for D_Q : 0.45 > u+O⁵ > 0.10; for D_Y : 0.50 > u+O⁵ > -0.10.

202

Table IX.3. Nonlinear specifications of the relation between growth, competitiveness, structure and openness (n=652)

Dep var	No	a	Ц	5	8	7	R'
Dq	i	fixed to 1	-0.65 (4.72 **8)	0.08 (0.68)	1.88 (5.54 *")	0.016 (7.20 •")	0.06
*D_B*	ii	0.41 (2.97 ***)	fixed to 0	0.72 (1.31)	1.87 (5.52 "*)	0.016 (7.22)	0.06
Dq	iii	fixed to 1	fixed to 0	1.03 (3.43 ***)	1.80 (5.24 ***)	0.014 (6.30 "*)	0.05
0,	iv	fixed to 1	-0.69 (7.20 •••)	0.18 (1.78 ***)	0.49 (2.13 ")	0.007 (4.75 *")	0.02
*D_r*	V	0.25 (2.77 ***)	fixed to 0	3.55 (5.27 *")	0.52 (2.25 •*)	0.007 (4.87 •••)	0.04
Dy	vi	fixed to 1	fixed to 0	1.42 (5.68 «*)	0.41 (1.72 *)	0.004 (2.93 ***)	0.03

Thus, the estimated coefficients for manufacturing yield values of the regression slope in the correct range, which holds to a lesser extent for total GDP. Although part of the estimated range is smaller than zero, keeping standard errors of the estimated coefficients in mind, these values are quite good.

The results of variants (ii) and (v) indicate that for GDP growth rate differentials, the linear (in the parameters) forms (vi) and (iix) in Table IX.2 are more restrictive than necessary. These equations yield significant parameters, which are different from those obtained in linear regressions. However, calculating the maximum value for the slope of the в term yields values around six for the GDP equation, and 0.8 for the manufacturing variant. The value for GDP is quite high from a theoretical point of view. Equations (iii) and (vi) bring the maximum values for the slopes of 0 closer to each other. For these equations, in which all parameters are significant, the values are 3.7 (GDP) and 2.5 (manufacturing). Still, this is quite high, so that one should interpret equations (ii), (iii), (vi) and (vii) as being not very relevant to the rightmost tail of the distribution of O.

The exact relationship between competitiveness, openness and growth as described by the first nonlinear equation is explained in Figures IX.la and IX.lb. The other nonlinear equations, as well as variants (ii), (iv), (vi) and (viii) in the linear estimates, yield similar 'landscapes', but they are somewhat less steep (and nonlinear) in the О dimension. In the figures, the 3-dimensional function described is projected on a 2-dimensional space, using the relevant ranges for the competitiveness and openness variable, and the estimated parameters for the manufacturing output growth variant of the equation. The structural part of the equation, as well as the constant, have been set to zero.

203

The two different figures project the same function in the 2-dimensional space, each taking a different viewpoint. Growth rate differentials are measured on the Z-axis (vertical), the X-axis (stretching from the bottom-left to the top-right corner) measures competitiveness, and the Y-axis (stretching from top-left to bottom-right) measures openness. For growth and competitiveness, the middle of the respective axes represent the point zero. For the Y-axis, the middle point corresponds with a value of the openness variable of around 125%. Moving to the right on the X-axis means a higher value of competitiveness, while moving to the left on the Y-axis corresponds to higher openness. Dark-coloured surfaces represent the top of the projected plane, and light shades correspond to the bottom of the plane.

In order to interpret the form of the plane, it is useful to start by imagining a situation in which openness does not matter, and the relation between competitiveness and growth is linear. This is the case in equations (i) in Tables IX.l and IX.2. Here, one could graph this relation in a 2-dimensional space. However, if a 3-dimensional space was used, there would be no variation along the third dimension, and the figure would simply look like an uphill road that can be crossed without gaining or losing height. Riding the road with a bicycle, however, would lead to gain/loss of height. At some point (halfway, at the point zero on the competitiveness axis), one would reach a point where the height corresponds to a zero growth rate differential.

Bearing this situation in mind, it is easy to see what would happen if the influence of openness is taken into account in as in the different variants of equation (IX.9). Now, each slice of the road (along its 'direction of the traffic') has a 'personal' steepness, which means that if one drives closer to one side of the road, the steepness varies. In fact, if one drives on the outer rightmost edge (corresponding to a closed economy), the road is completely flat. This slice of the road corresponds to the X-axis. The more one moves to the left, the steeper the surface becomes. This interpretation is evident from Figure IX.la.

Another way of saying the same thing is the following. At the maximum of openness (the farthest possible point on the Y-axis from the origin), the plane cuts the horizontal plane for which growth (Z) is zero with a fairly large slope (close to 0.7). From that slice on, the slices closer to the origin are curled towards the X-axis. For negative values of competitiveness (X-axis), the plane curls to the X-axis from below (negative values on the Z-axis), and for positive competitiveness it curls to the X-axis from above. This interpretation is more evident from the viewpoint taken in Figure IX.lb.

The simple, but important, economic interpretation of these figures is that competitiveness only matters when the economy is open enough. Economies actively taking part in world trade are more sensitive to differences in competitiveness than less open economies. To put it another way, it is not beneficial to have an open economy unless the domestic economy is competitive.

205

9.2. Catching Up: A Detailed Look at the Asian NICs

After this general interpretation of the relation between trade, competitiveness and growth, this section will take a closer look at some of the countries in the sample in order to see to what extent their growth pattern can be explained by the approach taken. The aim of this is to go beyond the general nature of the regressions in the previous section, and explore the data used there, as well as some additional data, for the consequences of the general framework derived from the model in Chapter 7.

The empirical overview in Chapter 4 has indicated that the NICs are the countries which have achieved the most spectacular growth performance in the period under consideration. But even within this group, there are considerable differences. Although the data are not actually documented here, it is a well-known fact that the Asian NICs³ are most remarkable. As will become apparent below, these countries have achieved very high growth rates over the previous period, which is the reason why they are sometimes called the Dynamic Asian Economies (DAEs). Thus, these countries seem to be good candidates for the case study approach adopted in this section. The USA (representing the economic and technological leader at the outset of the period) and Japan (the early example of catching up, and by now an economic and technological leader, especially in the Asian region) will also be considered as benchmarks.

Table IX.4 summarizes the growth performance of these countries. The table shows that at the outset of the period, the USA, as the technological and economic leader, was realizing a small positive growth rate differential. In manufacturing, most of the Asian economies were still falling behind, with Japan as a clear and Korea and Malaysia as less clear exceptions. For GDP, the growth rate differentials for the Asian economies were more on the positive side. Thus, Japan emerged as a regional leader in terms of growth rates and per capita income (not documented) as early as the 1960s. After 1965, the USA economy slowed down, and mostly achieved negative growth rate differentials. The Asian catching-up process set off in this period, and only came to a standstill in Japan in the most recent period. The other Asian economies, especially Korea, continued to grow very rapidly, both with regard to GDP and manufacturing, with occasional exceptions.

³ In this thesis: Hong Kong, (South) Korea, Malaysia, Singapore, Thailand, The Philippines. 206

Table IX.4. Growth performance of Asian NICs and technological leaders, 1963-1987

Country	1963-1965	1965-1970	1970-1975	1975-1980	1980-1985	1985-1987

Philippines	-2.95	■1.43	5.07	7.24	14.97	10.61
Malaysia	0.52	3.49	3.11	4.79	3.65	3.50
Thailand	-0.54	1.68	5.32	4.84	1.30	-4.82
Korea	1.17	16.40	14.48	14.31	7.04	11.91
Hong Kong	-1.33	3.12	4.37	6.42	4.04	6.48
Singapore	-2.19	6.27	-3.63	4.28	0.77	-1.01
Japan	2.50	6.22	1.23	1.19	2.01	-0.90
USA	0.72	-1.78	-1.39	0.05	-0.19	0.13
D,
Philippines	-2.08	0.54	2.43	2.23	-1.72	-1.99
Malaysia	0.15	1.00	4.31	4.59	3.52	-6.82
Thailand	1.85	2.71	1.62	4.26	3.05	0.73
Korea	0.10	5.91	5.77	2.85	2.25	5.64
Hong Kong	6.44	3.33	3.28	6.69	4.49	4.89
Singapore	-5.66	6.35	6.32	4.03	4.69	-1.18
Japan	3.20	5.25	2.03	1.23	1.13	0.22
USA	0.07	-1.34	-1.85	41.69	0.24	0.15

How can this growth pattern be explained? Bearing the results of the regressions in the previous section in mind, the present section explores the trends for the USA and Asian economies in more detail. The first factor that will be examined is competitiveness (€>). In Figure IX.2a, the competitiveness profiles of the USA and Japan are presented. The figure gives the percentage point contribution of wage rate competitiveness to the total on the horizontal axis, and its productivity and scale counterpart on the vertical axis. The solid line going from the upper left corner to the bottom right corner makes the distinction between negative (left) and positive (right) total competitiveness. The dotted lines divide the 2-dimensional space in parts that correspond to different sources of competitiveness. Japan starts as a technologically backward country, which is still competitive due to its low wage rate. The USA starts as a highly competitive country with regard to technology, but lags behind in the wage rate dimension. Some of the reasons why the USA's total competitive lag did not materialize in a larger negative growth rate differential than that in Table IX.4 will become apparent below. The catching-up process of Japan is made visible through its constant upward movement in the diagram. However, at the same time, Japan moves slowly to the left, indicating its loss in the wage rate dimension of competitiveness. The USA shows a movement in the opposite direction. The catching-up process in the rest of the sample makes

207

To sum up, Figures IX.2a - IX.2c illustrate one important source of the large positive growth rate differentials of the Asian economies, in the form of their high competitiveness. Compared to the other countries in the sample (not documented) which mostly move around the origin, or the solid line, competitiveness in these countries is very high. However, there are a number of countries for which the match between competitiveness in Figures IX.2 and growth rate differentials in Table IX.4 is not particularly good for some periods (with Korea in the early years being the most prominent example). This means that there must be additional factors explaining these countries' growth performance.

Table IX.5 gives two possible sources. First, the table shows the openness coefficients for the countries under consideration. The bottom line (sample mean) of the first half of the table shows the increasing internationalization of the world economy over the 1970s in the form of increased world trade. Over the 1980s, the trend in О is downward again, which does not necessarily indicate a decrease in internationalization. The presence of multinational companies might bias the particular statistic used here. Nevertheless, the numbers show the relatively large importance of domestic growth factors for the large economies of the USA and Japan. These two countries have a value of О clearly below the sample mean, which indicates the relatively small importance of competitiveness in this context. Both countries, however, have an increasing trend in O, in line with the world trend. The other Asian NICs can be divided into two groups. One group (Hong Kong, Singapore) is highly dependent on international trade, and achieves values of О far above the sample mean (in the case of Singapore even extremely high). The other group (The Philippines, Thailand, Korea) clearly has a value of О below the sample mean. Malaysia lies somewhat in between, with values of О around the mean. Korea starts at a very low level of O, but moves to a more open economy throughout the 1970s and 1980s. This shows that the logic of completely export-based growth is only valid for a limited number of the Asian 'tigers'. It also explains some of the mismatches between growth in Table IX.4 and competitiveness in Figures IX.2a ^J IX.2c.

One of the domestic sources of growth that has not been taken into account yet, is capital accumulation, which is presented in the bottom half of Table IX.5. Regarding the two leaders, USA and Japan, the well-known pattern is reproduced here. The USA has an investment ratio which is constant over most of the 1960-1980 period, but is well below the mean. Japan's investment intensity is much more volatile, reaching a peak in the mid-1970s, but is well above the sample mean for the whole period. The other Asian economies can be subdivided into two groups. The first group (Korea, Singapore, Malaysia) are following Japan, with investment intensities around average in the early 1960s, and above average throughout the 1970s and 1980s. The other group (The Philippines, Hong Kong) achieves around average investment intensities for the whole period. However, considering that the level of development in these countries is lower than in the OECD countries, one might interpret this as relatively high. Thailand is an exception to the high investment intensity in Asia, with levels well below those of the USA.

Table IX.5. Openness and investment Intensity, Asian NICs and technological leaders, 1963-1987

Country	1964-1965	1965-1970	1970-1975	1975-1980	1980-1985	1985-1987
0
Philippines	14.7	14.7	14.6	17.4	16.1	13.3
Malaysia	39.9	36.4	38.4	48.1	47.3	45.4
Thailand	12.5	12.6	14.0	18.1	17.8	16.4
Korea	7.5	12.9	22.6	39.3	46.2	42.9
Hong Kong	91.8	96.8	119.3	139.9	131.6	138.9
Singapore	141.5	135.2	159.6	na	na	na
Japan	12.8	13.8	21.1	30.5	32.8	31.5
USA	10.9	11.7	15.5	20.9	22.7	21.4
Sample mean	30.9	31.1	40.4	53.3	50.3	49.3
К
Philippines	18.3	19.2	19.4	23.9	20.2	12.6
Malaysia	22.4	23.2	28.2	31.1	37.1	32.0
Thailand	12.9	16.1	16.6	16.1	14.9	13.7
Korea	13.2	22.9	27.4	30.9	28.1	28.3
Hong Kong	25.8	19.4	19.5	22.3	21.4	18.3
Singapore	18.1	23.5	35.3	34.7	40.0	na
Japan	27.8	31.7	35.9	32.3	29.2	28.8
USA	16.8	16.6	16.8	17.3	17.6	18.8
Sample mean	23.6	23.9	25.3	24.6	23.0	na

211

Finally, the degree of specialization is investigated- The regressions in the previous section showed that this is a very strong factor explaining growth. As explained above, there are two sides to specialization, which can be broadly defined as demand side induced and supply side induced. Before separating these two, Figures IX.3a and IX.3b present the degree of specialization for the countries in the sample. The figures divide the countries into two groups, one with a high degree, and the other with a low degree of specialization. This division corresponds closely to the one made in the case of openness, which illustrates the relation between openness and specialization. The first group, in which the leaders Japan and the USA are found besides Korea and the Philippines, all have starting values of the variable И below the sample mean. The second group has starting values above the sample mean. In Figure IX.3a, note the trends for Japan and Korea. In the case of Japan, the increased openness over the 1970s and 1980s has also lead to increased specialization. For Korea, however, increased openness has gone together with decreased specialization (except for the period since 1985). This indicates the very broad range of the Korean competitive manufacturing activities.

In Figure ГХ.ЗЬ, note that all series show a decreasing trend over time. These countries, which all started at very high levels of specialization, are becoming less specialized over time. The sample mean is also decreasing, which shows that this is indeed a trend that is not limited to Asia. In light of the increasing trend in openness that was observed earlier, this is surprising, since one would expect increasing openness to go hand in hand with increasing specialization.

One possible explanation for this might be the dynamics of sectoral shares at the world level. In line with some of the theory on diffusion of innovations, long waves and technological paradigms, the share of new technologies in total (world) production has been rising since the late 1970s (see the discussion in Chapter 3). Reshaping the division of labour at the world level on the basis of this new sectoral division of labour is a process which erodes the old division of labour, and thus leads to a decreasing trend in specialization initially. This is consistent with the rise in specialization in some of the countries in Figures IX.3a and IX.3b since the beginning of the 1980s.

Table IX.6 investigates this hypothesis in more detail for the DAEs. The table illustrates the dynamics of specialization patterns in a nonquantitative way. The underlying statistics for this table are revealed comparative advantages (RCA), calculated on the basis of production statistics. The table gives the top-3 sectors for different criteria for different periods, which have been chosen on the basis of the movement in И in Figures IX.3a and IX.3b.

In the early 1960s, all Asian economies (including Japan) were specialized in low-tech sectors. Tobacco, wood and textiles were among the sectors with the highest RCAs. The USA, on the other hand, were specialized in instruments, transport equipment (mostly cars and aircraft), and refined oil.

The USA economy's specialization pattern remained largely constant, with the high-tech sectors at the forefront. However, in line with the Japanese competition in the automobile industry, transport equipment vanished from the RCA top-3 in

213

1987. Japan, on the other hand, showed major changes in its specialization pattern. The sectors which had been in the RCA top-3 in 1963 realized major losses in RCA (see the line top-3 changes -). The top-3 changes + sectors (instruments, electrical machinery and iron and steel) increased their RCA to such an extent that they became the top-3 sectors in 1987. Thus, the picture of Japan as a country switching from low-tech to high-tech production that was clear from Figure IX.2a, is confirmed by the dynamics of its specialization pattern.

Table IX.6. Specialization patterns in a dynamic perspective, Asian NICs and technological leaders, 1963-1987

Country	T,	T₂	T₃
Philippines	1965	1975	1987
ТорЗ +	Tobacco (9.1) Other chem (2.3) Paper (2.0)	Tobacco (7.4) Refined oil (2.8) Other chem (2.4)	Tobacco (7.1) Beverages (2.0) Apparel (2.0)
Top 3 -	Instruments (0.01) Misc C&O (0.07) Iron & steel (0.1)	Misc C&O (0.03) Instruments (0.04) Non-el mach (0.2)	Instruments (0.02) Non-fer met (0.02) Misc C&O (0.06)
Top 3 changes +		Refined oil (1.9) Printing (0.4) Iron & steel (0.3)	Fab met (1.6) Apparel (1.4) Beverages (1.1)
Top 3 changes -		Tobacco (1.7) Paper (1.3) Pottery (0.9)	Refined oil (2.0) Other chem (1.1) Wood (0.5)
Malaysia		1963	1987
Top 3+		Rubber (8.1) Wood (3.6) Tobacco (3.0)	Rubber (8.2) Wood (2.6) El mach (2.1)
ТорЗ-		Misc C&O (0.04) Transport (0.06) Instruments (0.06)	Misc C&O (0.07) Leather (0.1) Instruments (0.2)
Top 3 changes +			El mach (1.3) Apparel (0.4) Fab met (0.4)
Top 3 changes -			Tobacco (1.2) Wood (1.0) Food (0.6)
Thailand		1963	1987
ТорЗ +		Tobacco (3.6) Oth man (3.2) Food (3.1)	Apparel (3.2) Tobacco (3.2) Rubber (2.9)
ТорЗ-		Leather (0.09) Transport (0.1) Ind chem (0.1)	Printing (0.2) Non-el mach (0.2) Instruments (0.2)
Continued on next page-

Top 3 changes +			Textiles (1.2) Footwear (1.0)
Top 3 changes -			Food (1.2) Wood fum (0.5) Tobacco (0.4)
Korea		1970	1987
ТорЗ +	Tobacco (8.8) Pottery (5.0) Rubber (4.0)	Tobacco (6.8) Oth man (2.5) Refined oil (2.5)	Leather (3.1) Rubber (2.4) Textiles (2.3)
ТорЗ-	Transport (0.2) Leather (0.2) El mach (0.3)	Leather (0.1) Fab met (0.2) Transport (0.2)	Printing (0.2) Wood him (0.5) Fab met (0.5)
Top 3 changes +		Plastic (0.9) Refined oil (0.9) Ind chem (0.8)	Leather (3.0) El mach (1.5) Apparel (1.4)
Top 3 changes -		Pottery (3.6) Rubber (2.3) Tobacco (2.0)	Tobacco (4.5) Refined oil (1.6) Wood (1.5)
Hong Kong	1963	1976	1987
Top 3 +	Apparel (12.0) Textiles (4.9) Plastic (4.4)	Apparel (12.9) Textiles (4.9) Plastic (3.2)	Apparel (11.2) Instruments (6.1) Textiles (4.2)
ТорЗ-	Ind chem (0.1) Transport (0.1) Non-el mach (0.1)	Ind chem (0.1) Transport (0.1) Iron & steel (0.1)	Iron & Steel (0.1) Transport (0.1) Non-fer met (0.2)
Top 3 changes +		Apparel (0.8) El mach (0.7) Instruments (0.7)	Instruments (4.4) Oth man (2.4) Tobacco (0.8)
Top 3 changes -		Plastic (1.2) Printing (0.8) Oth man (3.5)	Apparel (1.7) El mach (1.3) Textiles (0.7)
Singapore		1968	1986
Top3 +		Rubber (7.2) Refined oil (5.7) Wood (2.7)	Refined oil (5.6) El mach (2.1) Rubber (1.9)
ТорЗ-		Paper (0.1) Pottery (0.1) Iron & steel (0.1)	Pottery (0.1) Glass (0.1) Textiles (0.1)
Top 3 changes +			El mach (1.8) Apparel (0.7) Ind chem (0.4)
Top 3 changes -			Rubber (5.3) Wood (2.4) Footwear (1.0)
Continued on next page.-.

215

Japan	1963	1987
Top 3 +	Wood (2.6) Oth man (2.2) Plastic (2.0)	El mach (2.0) Iron & steel (1.4) Instruments (1.2)
Top 3-	Footwear (0.3) Instruments (0.5) Refined oil (0.6)	Footwear (0.4) Apparel (0.5) Wood fum (0.6)
Top 3 changes +		El mach (1.1) Instruments (0.7) Iron & steel (0.4)
Top 3 changes -		Wood (1.6) Wood furn (1.2) Oth man (1.0)
USA	1963	1987
Top3+	Instruments (1.5) Refined oil (1.5) Transport (1.3)	Instruments (1.4) Printing (1.3) Refined oil (1.3)
Top 3-	Pottery (0.4) Plastic (0.4) Textiles (0.6)	Footwear (0.4) Pottery (0.4) Leather (0.5)
Top 3 changes +	■■ v------	Plastic (0.5) Wood furn (0.3) Printing (0.3)
Top 3 changes -		Iron & steel (0.4) Footwear (0.4) El mach (0.3)

The other Asian economies are less dynamic in this respect. Still, a number of them (Malaysia, Hong Kong, Singapore) manage to become specialized in at least one high-tech sector (electrical machinery or instruments). In Korea, electrical machinery is becoming increasingly important, as is clear from the top-3 changes + row, but in 1987, this sector did not rank among the top-3 RCA yet. Thus, all the economies which have shown an increasing trend in specialization over the last years, are becoming more specialized in at least one high-tech sector. This illustrated the value of the argument on the reshaping of the international division of labour.

Another point that is worth mentioning from Table IX.6 is the importance of natural resource-based industries such as refined oil, rubber and wood for some of the Asian economies. Being tied to one particular location, these sectors would be expected to have a high degree of specialization. Nevertheless, the fact that most Asian economies have been able to go beyond the exclusive exploitation of these natural resources, and specialize in other sectors as well, is indicative of their dynamic power. Take Korea, which was specialized in the natural resource based industries rubber and pottery in 1963. At that time, electrical machinery and transport were very weak. Over the 1960s, Korea experienced a wave of specialization, which led to a strong position in other manufacturing, and refined oil as a new, but more important, resource-based industry. After 1970, the Korean

216

economy was once again reshaped, this time resulting in a net decline of specialization. During this period, a number of its weaknesses (leather, electrical machinery) in 1963 developed into strengths by 1987.

There are two other countries for which wave-like patterns in specialization were found. The Philippines switched to an oil-based economy during the 1960s and early 1970s. Subsequently, it became specialized in low-tech sectors such as beverages and apparel. Note also the constant importance of tobacco over the total period. Hong Kong showed specialization waves too, but only over the most recent period was this accompanied by a real structural change in the sense that instruments became more and more important. Over the total period, this country is the best example of the importance of textiles(-related) industry in the Asian countries.

Thus, this table shows the supply side effects and causes of specialization. Table IX.7 shows the demand side effects related to specialization. The table gives the average value of the variable S over different subperiods. Overall, the value of S (compared to growth rate differentials and/or competitiveness) is small. However, as the regressions showed, the importance is still significant, The table shows that the two leaders (Japan, USA) mostly stay on the positive (including 0) side (Japan especially since it has become a technological leader). The other countries are mostly on the negative side, except for Hong Kong for the post-1960s period, and the other countries for the period around the first oil crisis. This indicates that the catching up of the Asian NICs is not due to the demand effects of structural differences. If anything, this effect was beneficial to the economic and technological leaders.

Table IX.7. Growth due to the demand side effect of structural differences, Asian NICs and Technological Leaders, in percentage points

Country	1963-1965	1965-1970	1970-1975	1975-1980	1980-1985	1985-1987
Philippines	-0.8	-0.5	0.1	-0.1	-0.4	-0.5
Malaysia	-1.3	-0.5	0.2	0.1	0.2	-0.2
Thailand	-1.7	-1.3	0	0	-0.5	-0.4
Korea	-0.83	-0.25	0.1	-0.2	-0.4	-0.3
Hong Kong	-1.2	-0.5	0.2	0.1	0.2	0.1
Singapore	-0.4	0.5	0.4	-0.2	-0.6	-0.5
Japan	-0.1	0	0	0	0.3	0.2
USA	0.2	0.1	0	0	0	0.1

217

Summarizing the results, it is clear that there is not a general strategy for successful catching up, even when attention is limited to a small group of countries such as the Asian NICs. Table IX.8 summarizes the findings of this section in a qualitative overview of some the factors stimulating growth in the Asian economies. Overall low wages seem to (have) be(en) a very strong factor stimulating growth in Asia. In addition to this, investment plays a (major) role in all countries, except Thailand. More recently, growth in high-tech industries has been a factor contributing in a positive way. Overall, the Asian economies do not seem to be specialized in sectors with very high (world) income elasticity. Other factors, such as specialization and openness vary between the different Asian economies. Some of the countries have benefited from these factors, while others have had a weaker position in this respect.

Table IX.8. Factors explaining growth in the Asian NICs

Factor	Period	Philippines	Malaysia	Thailand	Korea	Hong Kong	Singap ore	Japa n
Low wages	early		++	++	+	++	+	++
Low wages	late	++		++	++	++	++	0
Technology	early	-	+	-	-	-	+	-
Technology	late	-	у	-	0	+	+	+
Natural resource based industries	early	+	++	0	0	0	++	+
Natural resource based industries	late	0	++	+	0	0	++	-
High-tech industries	early	0	-	-	0	-	0	0
High-tech industries	late	0	+	-	+	+	+	++
Growth from trade	early	0	+	0	0	++	++	0
Growth from trade	late	0	+	0	+	++	++	+
Specialization	early	о ^;	+	+	0	++	++	-
Specialization	late	0	0	0	0		++	0
Income elasticity of demand	early	-	-	-	-	-	0	+
Income elasticity of demand	late	-	0	-	-	+	-	0
Investment	early	+	+	-	+	+	+	++
Investment	late	+	++	-	++	+	++	++

218

9.3. Conclusions

The tests performed in section 9.1 show the general relevance of the approach to growth in the model presented in Chapter 7. The results show the importance of the evolutionary argument that differences in economic structure and differences in competitiveness are related to growth rate differentials. Thus, the model and the tests based on it have shown to have some value in explaining the first stylized fact in Chapter 4. It has also been shown that the second and third stylized fact are significant factors in explaining the first one. However, since the explanatory power of the regression is low, it also appears that one needs to consider more factors determining competitiveness than the limited set used here and in the previous chapter. To do so, further research must be carried out.

The case study approach adopted in section 9.2 also shows this. The growth pattern of the Asian NICs can be explained by looking at the variables in the regressions, but there are also parts which cannot be explained directly by this method. Moreover, the case study approach shows that variety is a very important concept in explaining growth. As the evolutionary logic stresses, there is not a generally valid strategy for catching up. Each of the countries considered has its own specific factors fostering growth.

This leads to a specific way of looking at growth that is quite different from the mainstream models outlined in Chapter 2, which treat growth as a balanced phenomenon, having a gradual nature. The open economy evolutionary logic, on the contrary, looks at growth as induced by changes in trade and the selection environment, which can be quite sudden and unexpected. Thus, the growth paths of the economies are subject to sudden shocks and trend reversals.

However, the concluding section of this last empirical chapter is not the place to discuss this issue in depth. This will be done in the last chapter, which will summarize the main arguments and conclusions.

219