Openness, Capital Accumulation and EconOmic Growth: The Evidence from Sub-Saharan Africa

Olasupo Akano

Abstract: Relative to most of their Asian counterparts, Sub-Saharan African economies have generally performed poorly in meeting their growth objectives since the 1960s. Using individual country time series and panel regressions, the paper investigates the role of the growth of capital stock per worker and openness in the economic growth experience of eleven Sub-Saharan African countries since the mid-60s. While the openness variable was found to have a strong positive effect on the observed pattern of economic growth, the capital-labour ratio paradoxically seems to have made a significant negative contribution to the growth process of the countries studied.

1. Introduction

The traditional neo-classical growth function postulates capital stock per worker as the critical determinant of aggregate output growth among nations (Solow 1956; Swan 1956). More recent empirical evidence seems to suggest that while physical capital accumulation is a necessary prerequisite for output growth, it might not be sufficient. The on-going controversy among theorists on the potential role of trade policy orientation in long-run growth underscores the seemingly limited scope of traditional theory. On the one hand, there is a growing body of time-series and cross-country empirical literature suggesting that international openness has a positive effect on economic growth (Romer 1986; Lucas 1988; Dollar 1992; Edwards 1992; Harrison 1995). On the other hand, there is an equally influential body of research that casts doubt on the hypothesised positive relationship between openness and growth (Kruger 1994; Rodrik 1995).

Within the context of the hypothesis, some writers suggest that Africa's generally poor growth record, particularly since the early 1970s, was most probably due to, among other factors, the inward-looking and largely protectionist trade policy regimes adopted by most Sub-Saharan African countries (Easterly and Levine 1997, 1998). One of the main pillars of the World Bank/IMF economic reform programs in Africa since the early 1980s has been trade liberalisation. The underlying premise of these policy reform regimes was that a systematic dismantling of official controls on external trade would facilitate the growth of total factor productivity across the continent (Obstfeld and Rogoff 1996). Indeed, the sharp contrast in the macroeconomic performance of south and south-eastern Asian countries vis-à-vis their African counterparts has been ascribed mainly to the divergent trade policy orientations of the two regions (Greenaway and Nam 1988; Dollar 1992).

The objective of this paper is to examine the time-series role of changes in international openness and capital stock per worker in the growth experience of a cross-section of Sub-Saharan African countries. The rest of the paper has been organised as follows. In Section 2, the theoretical basis of the hypothesised positive relationship between economic growth, capital accumulation and openness is discussed. This is followed in Section 3 by the specification of and rationale for the dynamic growth model adopted in the paper. Measurement of the variables and issues relating to sample selection are addressed in Section 4 while the estimated results are discussed in Section 5. The final section contains the conclusions of the paper.

2. Theoretical Framework

The standard neo-classical production function incorporating technological progress specifies aggregate output-input relation as:

Y_(t) = F(A_(t) K_(t), B_(t)L_(t)) ... ... ... (1)

such that

A_(t) = B_(t) = m ... ... ... (2)

A_(t) B_(t)

where Y denotes total output, K is the capital stock, L is the labour force, and the multiplicative constants A and B, functions of time, represent factor productivity or technology coefficients. Assuming constant returns to scale, (1) and (2) imply that technical progress is equally capital and labour-augmenting at the constant proportional rate m and 910 may be re-expressed as:

Y_(t) = A_(t)F(K_(t), L_(t)) ... ... (3)

There have over the years been two fundamental objections to the traditional growth theory as summarised in (3). First, contrary to (2), the rates of change of technical progress are not constant; and secondly, they are not exogenous to the economic system. It is contended, within the framework of the new growth theories, that the rate of change of technological advancement is endogenous to specific economies and are characteristically determined by a combination of internal and external factors. Expressing the variables in (3) in per capita terms, we have

y_t = A_tf(_kt) ... ... (4)

where y_t = Y_t/L_t and _kt = K_t/L_t

On the assumption that the production function (4) satisfies the Inada conditions (Inada 1963), traditional neo-classical growth theory postulates output per worker as a positive function of the capital intensity of production (k) at any given period of time. More recent approaches to the dynamics of economic growth, which form the foundation of the new growth theories (Romer 1986; Lucas 1988), centres around the presumption that technical progress is not only endogenous to an economic system, it also varies from period to period.

The contention of the new growth theories therefore is that there are inter-country differences in the rates of technological innovation and/or imitation. It seems intuitively plausible, from the viewpoint of the new theories, that these differences explain, among other factors, international disparities in total factor productivity growth at any period of time.

Following Edwards (1992), this paper maintains that technological progress can emanate from two sources: domestic and foreign. While internally induced technical improvements are often embodied in a country's exports, technical innovations from abroad can similarly be embodied in the country's imports (particularly of capital goods and processes). Thus, the avenue of external trade could provide the basic transmission mechanism of endogenous technology. If this assumption is valid, as Lewis (1995) asserts, then the greater the degree of openness of an economy, the greater the country's access to new ideas from the technologically more advanced countries.¹ By contrast, the rate at which relatively more closed economies can absorb the wide range of productivity-enhancing technical innovations would be expected to be slower than that of the more open economies.

3. The Model and Methodology

If the traditional neo-classical growth model specified in (4) is expanded to incorporate the openness variable, we obtain

y^__t = f (k_t, p_t )  y_t k_t p_t 

or

G_t = G(gk_t, gp_t) ... ... ... (5)

where

Gt = y_t, gk_t = k_t, and gp_t = p_t , yt kt p_t

such that the endogenous technology variable A_t is proxied by the openness variable p_t, given by

p_t = (X_t = M_t)/Y_t

where X_t denotes exports, M_t is imports and Y_t is GDP in real terms.

The basic model (5) postulated in this paper has been rationalised in three related ways. The first version of the basic model is an autoregressive distributed lag (ADL) specification given by

i=l j=0 h=0

where

G_t denotes the proportionate rate of change of real aggregate output in period t; gk_t is proportionate rate of change of capital stock per worker; gp_t is rate of change of openness of the economy; __ __s, __s and __s are constants; and __t is the error term.

The dynamic specification of the growth model is premised on the logic that economic growth is an accumulative process of change and transformation. Thus, the autoregressive character of (6) is founded on the contention that output growth in a given period might be determined not solely by other substantive variables (e.g., capital stock) but also by levels of growth achieved in proceeding periods of time. Thus, in general terms, we have the autoregressive function

m

G_t = __o +_ __i LⁱG_t + v_t ... ... ... ... (7)

i=l

m

where _ __i _ 1

i=l

The basic assumption underlying an autoregressive growth model as in (7) is that output growth rates in previous periods provide the foundation for current growth attainments. It is, of course, assumed that the underlying macroeconomic policy regimes that influence the determination of real output growth are constant from period to period. It is also recognised that it is impossible to establish from the specification in (7) what the maximum lag length should be. The approach adopted in this paper was to select an arbitrary maximum lag length, (taking into account the requirements of a satisfactory degree of freedom).

Moreover, the distributed lag structure of the capital stock variable is based on the contention that capital accumulation is a long-term process, the dynamics of which are provided by the savings-investment cycle. Thus, observed growth rate of per capita output at any period is the result of the cumulative process of time-series changes in physical capital accumulation not just in the present but also over several periods in the past.

It is, of course, possible for a country to accumulate increasing quantities of physical capital without the stock being translated into appreciable positive output growth. This might be due to a variety of reasons. First, the human capital and technological base of the economy might be weak. In such a situation, vintage methods of production that inhibit the growth of total factor productivity might permeate the economic system. Second, it might be that, in spite of an appreciable stock of capital per worker, the process of economic growth was hindered by the kind of official policies adopted to regulate economic activity over the long term. Aside from "wrong" application of the spectrum of domestic monetary and fiscal policies, there are also the kind of external trade policies adopted by the government to regulate the flow of international transactions. With regard to the latter group of obstacles to growth, there is a growing body of evidence that seems to suggest that there is a predictable and systematic long-term relationship between a country's external trade orientation and its pattern of growth. It is argued that countries that pursue enduring outward-looking trade policy regimes tend to grow faster than those with constricting trade policies (Dollar 1992; Edwards 1992; Sinha and Sinha 1998). The methodology of the paper is thus to use the ADL procedure to evaluate the relative impact of changes in capital stock and trade policy orientation on the observed pattern of economic growth in selected Sub-Saharan African countries since the mid-60s.

One of the major limitations of the ADL methodology specified above derives not from the basic concept itself but from the non-availability of long-range data on the relevant variables in most African countries. First, very few Sub-Saharan African countries have published data on the capital stock variable, and these are mostly of limited span. Second, most of these countries did not have any credible database until after the attainment of political independence in the 1960s.These observations boil down to the fact that the required large samples necessary for unbiased and consistent estimation of the parameters of the time-series regression for each country are lacking. The data limitation that plagues the postulated time-series and third versions of model (5) are based on the pooled time-series and cross-section procedure that is specified as follows:

G_it = _ + _gk_it + _ gp_it + __it, i = 1,2,...,11; t = 1,2,...,26 ... .... (8)

where the variables are as defined above and i and t denote countries and years, respectively. The panel approach is expected to overcome the problems of small-sample biases that could affect the reliability of the pure time-series mode (6).

4. Data Measurement and Country coverage

Eleven Sub-Saharan African countries for which required time-series data are available are covered in this paper. All the data used in the study are obtained from the Penn World Table, version 5.6, constructed by Summers and Heston (1991). The country coverage of the study is limited by the fact that the capital stock variable is available for only the Sub-Saharan African countries listed in table 1. The three broad groups of variables included in the dynamic model (see equation 6) are measured as follows:

Growth: The growth variable (G_t), the dependent variable, is defined as the rate of growth of the real GDP per capita. The time-series data covers the 1965-90 period for each country.

Capital stock/worker: The basic capital stock variable was measured as physical capital stock per worker. The data for this variable also covers the period 1965-90 for all the countries except Botswana, in respect of which data for 1987 and 1988 were estimated as five-year moving averages over the previous five years, and Swaziland, for which similar estimates were made for the period 1986-88.

Openness: The openness index used in the paper is total trade (export plus import) as a proportion of GDP, as provided in the Penn World Table. One of the fundamental weaknesses of this index is the negative effect of trade restrictions on a country's total trade. The relative strengths of official trade restrictions on a country's exports and imports are likely to have a more or less distorting impact on the numerical value of the index at any point in time. However, the openness index, as defined, might be expected to produce a fairly accurate reflection of a country's trade orientation if the strong assumption is made that trade barriers on a country's exports and imports produce equal and opposite effects, thereby cancelling each other out. It is however generally recognised that trade in natural resource products, especially agricultural commodity exports from African countries, are more subject to restrictive trade policies in the industrialised countries than manufactures (e.g., European Union's Common Agricultural Policy). It is also known that certain categories of manufactures, e.g., textiles, clothing and footwear, in which several developing countries have a certain measure of competitive edge over their foreign rivals, were excluded from the provisions of the Generalised System of Preferences (GSP) since the early 1970s. These two examples have the potential effect of lowering the numerical values of the openness index for some countries in the sample.

5. Analysis of Results

In view of the relatively small sample size (N=26) for each country, the paper postulates an ADL model specified in (6) such that n=m=2. The estimated time-series equation for each of the eleven countries using the OLS procedure is presented in table 1. In the first stage of the model reduction process, a joint test of statistical significance of each block of explanatory variables included in the general model was carried out. The regression results obtained in the process of the sequential exclusion of lagged output growth rates, growth rates of capital stock per worker, and the openness index are reported in tables 2- 4. The paper went further to test the statistical significance of each lag in the postulated general model. The estimated regression results for each stage in the model reduction process, the F version of the Lagrange Multiplier (LMF)² test statistic for each block of variables and for each lag was calculated in order to test the validity of excluding a specific group of variables or lag. Following the statistical validation of the variable/lag exclusions as a logical part of the model reduction process, the structure on the final reduced model for each country is summarised in table 7. These equations are finally re-estimated using the OLS procedure and the validity of all the variable/lag exclusions tested using the LMF test statistic. The final estimated reduced model is presented in table 8.

The Augmented Dickey-Fuller procedure was used to test the stationarity of each of the variables, first in the general specification and then in the reduced model for each country (table 8). The same test procedure was used to determine the appropriate data-generating process for each of the variables appearing in the final regression for each country. The results of these tests are summarised in tables 5.1-5.3. The tables show that the variables G_t, gk_t and gk_t-2 are stationary over the 1965-90 period.³ The last column of each of these tables presents the "appropriate" data-generating process (based on the Dickey-Fuller F-test) for the three variables. Except for the G_t variable, which exhibited both drift and trend in the Madagascar data, the DGP for all the three variables in the other countries exhibited no trend over the period.

The most visible and unexpected result form the final estimated regression equations presented in table 8 was the absence of the openness variables (gp_t-j, j = 0,1,2) from all the countries included in the sample. In the course of the model reduction process, the current as well as lagged values of the openness variable were found not to have made any significant contribution to the economic growth process in these countries over the 1965-90 period. This result is contrary to the run of more recent thinking in the new growth theories and empirics (Collier and Gunning 1999; Harrison 1995). Indeed, this result does not seem logical in view of the fact that African countries, on average, were more open (using the adopted definition) than most of the more developed countries over the period under reference.⁴ Moreover, the seemingly endemic current account deficits experienced by most of these countries since the 1970s appear to be a reflection of the relative dependence of the economies of Sub-Saharan African countries on foreign trade over the period. It would thus be expected that the long-term market dependence of the continent's industrial sectors on imported capital goods and raw materials would translate into improved access to productivity-enhancing technical innovations from the more advanced countries over the reference period.

The result of the panel regressions that are reported in tables 12 and 13 seems to reverse the apparently anomalous time-series regression result. However, while the openness variable was seen to make a significant positive contribution to economic growth in Africa over the period, the panel result simply confirms the country time-series findings that there was a significant negative correlation between growth and the capital-labour ratio over the same period. An examination of the results presented in tables 1, 2, and 4-6 will reveal that the capital-labour variable for most of the countries displayed negative signs. There was the initial suspicion that the problem of intercorrelation among the lagged capital intensity variables might be responsible for the seemingly contradictory signs. However, the panel result presented in table 12, where no lagged regressors were present, indicates that the negative sign attached to the capital intensity variable was apparently not caused by the problem of multicollinearity.⁵ The Hausman _k² values reported below in tables 12 and 13 indicate that the null hypothesis of zero country-specific effects could not be rejected at the 5% level. Hence, the Random Effects estimator is preferred to the Within (Fixed Effects) estimator in either case. While the theoretically inconsistent signs of the capital intensity variables in both the time-series and panel regressions could not be readily explained, the latter set of estimates seems to be more realistic in view of the fact that it reduces to relative insignificance the statistical problem of small-sample bias.

6. Conclusions

The objective of the paper has been to evaluate the role of openness and capital intensity in the growth process of selected African countries over the 1965-90 period. One of the main findings of the paper is that the openness variable made statistically significant positive contributions to the growth experience of the countries studied. Evidence from the literature (see Edwards 1992, 42) seems to support the contention that there is a positive relationship between openness and growth. However, contrary to theoretical exception, the paper found a statistically significant negative correlation between economic growth and the growth of capital-labour ratio in the countries included in the sample.

Table 1. Estimated ADL Model-Regression of growth on the growth of the capital stock and openness variables

	_	_1	_2	_o	_1	_2	_o	_1	_2	R2	F
Botswana	-2.831 (-1.4047)	-.0489 (-.3808)	-.0060 (-.0912)	-.0817 (-.7080)	.0248 (.6019)	1.3896 (32.37	.1116 (.8975)	.0208 (.3154)	-.0151 (-.1206)	0.9919	200.00
Cote d'Ivoire	12.706 (2.2180)	-.2733 (-.8485)	.2345 (.7189)	-.2381 (-1.090)	-.1519 (-2.260)	.8253 (12.386)	.1491 (.5117)	-.3005 (-1.0279)	.1159 (.4762)	0.9891	146.88
Kenya	20.380 (2.6713)	-.2195 (-1.0107)	-.0990 (-.6011)	.0527 (.3856)	-.2071 (-2.277)	.7497 (7.9977)	-.0419 (-.2471)	-.1074 (-.7865)	-.1975 (-1.3056)	0.9730	58.54
Madagascar	-91.5026 (-.8417)	.0310 (.1391)	.0294 (.1331)	.0887 (.4001)	.0685 (.8151)	1.0373 (12.357)	.0374 (.1742)	.0391 (.1833)	-.0229 (-.1060)	0.9926	217.44
Malawi	4.1240 (1.1284)	-.1988 (-.6873)	-.4403 (-1.6140)	.0112 (.0324)	-.1184 (-.9006)	.8109 (7.5509)	.0990 (.2758)	.4543 (1.2921)	-.0652 (-.1263)	0.9701	52.69
Mauritius	45.989 (1.0639)	.0519 (.1416)	1592 (.5056)	-.0632 (-.1788)	-.1543 (-.9549)	.8567 (5.1847)	-.2375 (-.4939)	-.3308 (-.8742)	-.1130 (-.2703)	0.9189	18.42
Nigeria	7.8557 (.4863)	.0948 (.3430)	-.0498 (-.1804)	-1.6405 (-2.8579)	-.0863 (-.5948)	.8952 (5.6269)	-.1849 (-.5018)	.0410 (.1115)	1.7705 (2.7337)	0.9468	28.91
Sierra Leone	7.1935 (1.0947)	-.0162 (-.0814)	.0092 (.0479)	-.1001 (-.6218)	-.0438 (-.8250)	.8929 (16.906)	.-0380 (-.1982)	-.0421 (-.2226)	.0870 (.4668)	0.9784	73.75
Swaziland	22.00 (.6135)	.0370 (-.2283)	-.0414 (-2578)	-.0707 (-.4558)	-.0819 (-.5833)	.9219 (6.6259)	-.0345 (-.2410)	-.0356) (-.2436)	-.0194 (-.1385)	0.9646	44.25
Zambia	-4.2660 (-.6110)	-.0146 (-.0621)	.0168 (.0937)	-.0903 (-.8800)	.0401 (.6091)	.9611 (15.403)	.0249 (.1075)	.0035 (.0188)	.1772 (1.0283)	0.9772	69.75
Zimbabwe	-14.829 (-.6616)	.0654 (.3212)	.0473 (.2577)	.1147 (.4608)	.0385 (.5651)	1.0529 (15.409)	-.0201 (-.1561)	-.0077 (-.0668)	-.0825 (-.3607)	0.9800	79.785

Model: G_t = _ + _₁G_t-1 + _₂G_t-2 + __ogk_t + _₁gk_t-1 + _₂gk_t-2 + __ogp_t + _₁gp_t-1 + _₂gp_t-2 + v_t

Note: Figures in parentheses beneath the estimated parameters refer to the Student's t-ratios

Table 2. Hypotheses: _₁ = _₂ = 0

	_	_o	_1	_2	_o	_1	_2	R2	F
Botswana	-2.7231 (-1.4895)	-.0855 (-.7941)	.0228 (.5980)	1.3878 (35.043)	.0659 (2.4246)	.0147 (.5452)	-.0210 (-.0944)	0.9918	304.020
Cote d'Ivoire	12.6156 (2.6721)	-.2205 (-1.0298)	-.1701 (-2.9807)	.8256 (12.946)	-.1029 (-1.6013)	-.0656 (-1.1108)	.1000 (.4232)	0.9878	202.42
Kenya	16.295 (2.5732)	.0597 (.4494)	-.1825 (-2.1985)	.7887 (9.3748)	-.1810 (-2.4639)	-.1475 (-2.0904)	-.1646 (-1.1518)	0.9705	82.19
Madagascar	-86.662 (-.8798)	.0853 (.4138)	.0650 (.8506)	1.0338 (13.539)	.0634 (.8499)	.0637 (.8535)	.-0229 (-.1137)	0.9926	333.60
Malawi	3.5231 (.9408)	-.0130 (-.0371)	-.0927 (-.7240)	.9025 (9.4080)	-.1018 (-.6506)	-.0770 (-.5358)	-.0721 (-.1373)	0.9633	65.70
Mauritius	45.6901 (1.1375)	-.0654 (-.1969)	-.1561 (-1.0323)	.8578 (5.5721)	-.1768 (-.9942)	-.1586 (-.9481)	-.110 (-.2814)	0.9172	27.69
Nigeria	7.5656 (.5082)	-1.6321 (-3.040)	-.0783 (-.5822)	.9024 (6.2278)	-.0785 (-.4976)	-.0133 (-.0854)	1.7537 (2.8989)	0.9462	43.94
Sierra Leone	7.1539 (1.1713)	-.0996 (-.6652)	-.0439 (-.8908)	.8934 (18.333)	-.0530 (-1.0312)	-.0332 (-.6336)	.0865 (.4988)	0.9784	113.39
Swaziland	16.135 (.5604)	-.0581 (-.4142)	-.0608 (-.5254)	.9433 (8.2777)	-.0472 (-.4880	-.0511 (-.5073)	-.0127 (-.0982)	0.9643	67.518
Zambia	-4.2291 (-.6832)	-.0910 (-.9536)	.0390 (.6826)	.9612 (17.454)	.0116 (.2206)	.0190 (.3372)	1780 (1.1109)	0.9772	107.20
Zimbabwe	-9.5684 (-.6084)	.0919 (.4089)	.0272 (.4901)	1.0403 (18.923)	.0160 (.4504)	.0163 (.4577)	-.0699 (-.3309)	0.9798	121.48

Note: Figures in parentheses beneath the estimated parameters refer to the t-ratios

Table 3. Hypotheses: __o = _₁ = _₂ = 0

	_	_1	_2	_o	_1	_2	R2	F
Botswana	26.021 (1.9901)	.3813 (.3356)	-.2431 (-.4121)	-.7551 (-.6977)	.1509 (.2556)	.1559 (.6541)	0.1942	0.7712
Cote d'Ivoire	36.818 (2.5837)	-.5547 (-.3435)	2.2064 (1.5917)	-.3103 (-.2102)	-2.3975 (-1.8704)	.2117 (.8132)	0.6480	5.8906
Kenya	46.2346 (3.7886)	-1.3194 (-1.6256)	-.0180 (-.0289)	.4274 (.6505)	-.4940 (-.9168)	-.1814 (-.9032)	0.4612	2.7395
Madagascar	621.56 (3.7930)	-.4786 (-.2572)	.0664 (.0361)	-.0202 (-.0114)	-.5253 (-.2976)	-.4346 (-2.1683)	0.3481	1.7085
Malawi	12.9097 (2.2240)	-.0528 (-.0710)	-2.2929 (-3.7280)	-.7154 (-.7488)	2.4623 (2.9376)	.1522 (.4995)	0.6924	7.2039
Mauritius	142.08 (3.4423)	.3849 (.4094)	.2349 (.2881)	-1.2152 (-1.0861)	-.8016 (-.8684)	-.3014 (-1.1541)	0.3286	1.5667
Nigeria	61.7894 (3.4962)	.540 (.6760)	.8763 (1.1322)	-1.4738 (-1.6263)	-.1.5641 (-1.6393)	-.0810 (-.2734)	0.4281	2.3954
Sierra Leone	54.7791 (2.7251)	-.3850 (-.3688)	.4853 (.4800)	-.1616 (-.1611)	-.7164 (-.7356)	-.0460 (-.1900)	0.2551	1.0960
Swaziland	126.35 (3.5584)	-.2582 (-.4302)	-.2767 (-.4596)	-.2557 (-.4817)	-.1874 (-.3537)	-.3314 (-1.6089)	0.2977	1.3562
Zambia	47.0131 (2.7654)	.2007 (.1802)	.7136 (.7898)	-.6780 (-.6570)	-.9806 (-1.1126)	-.0919 (-.4006)	0.2597	1.1226
Zimbabwe	159.75 (3.1936)	-1.1854 (-1.1670)	-.3817 (-.4284)	.5065 (.7336)	-.0192 (-.0310)	-.2235 (-1.3555)	0.2553	1.0970

Note: Figures in brackets beneath the estimated parameters refer to t-ratios

Table 4. Hypotheses: __o = _₁ = _ = 0

	_	_1	_2	_o	_1	_2	R2	F
Botswana	-2.2606 (-1.3038)	.0629 (2.2481)	.0089 (.3370)	-.0965 (-3.9386)	.0204 (.5446)	1.3819 (35.85)	0.9914	367.97
Cote d'Ivoire	14.0159 (2.7908)	-.1240 (-1.8218)	-.0862 (-1.3231)	-.1453 (-2.7255)	-.1817 (-3.1285)	.8176 (12.95)	0.9878	258.42
Kenya	14.976 (2.3214)	-.1990 (-2.239)	-.1516 (-1.8475)	-.0945 (-1.3117)	-.1530 (-1.9352)	.8090 (9.7751)	0.9683	97.80
Madagascar	-84.0011 (-.8922)	.06425 (.8605)	.06427 (-.8646)	.0613 (.8848)	.0627 (.8616)	1.0315 (14.185)	0.9925	425.79
Malawi	5.2171 (1.6653)	-.1499 (-1.5506)	-.1167 (-1.2076)	-.0928 (-1.1453)	-.1298 (-1.4516)	.8546 (9.3686)	0.9662	91.47
Mauritius	24.353 (.7994)	-.0763 (-.6226)	-.0609 (-.5096)	-.0962 (-.7834)	-.0936 (-.7305)	.9267 (7.2558)	0.9135	33.80
Nigeria	11.315 (.8235)	-.0857 (-.6664)	-.0698 (-.5391)	-.0954 (-.6969)	-.1100 (-.7504)	.9646 (6.832)	0.9150	34.46
Sierra Leone	6.7798 (1.1852)	-.0487 (-.9589)	-.0319 (-.6128)	-.0255 (-.5986)	-.0414 (-.8770)	.8957 (19.260)	0.9779	141.80
Swaziland	15.085 (.5763)	-.0497 (-.4899)	-.0545 (-.5199)	-.0664 (-.7150)	-.0568 (-.5347)	.9468 (8.9854)	0.9643	86.40
Zambia	-.4013 (-.0904)	-.001 (-.0190)	-.0090 (-.1861)	.010 (.3270)	.0070 (.1495)	.9390 (19.694)	0.9753	126.35
Zimbabwe	-12.0159 (-.6513)	.0308 (.5372)	.0294 (.5219)	.0250 (.6034)	.0322 (.5484)	1.0461 (17.742)	0.9798	155.29

Note: Figures in brackets beneath the estimated parameters refer to t-ratios

Table 5. Hypotheses: _1 = _1 = _1 = 0

	_	_2	_o	_2	_o	_2	R2	F
Botswana	-1.999 (-1.3330)	.0133 (.5033)	-.0810 (-.7662)	1.3779 (39.013)	.0630 (2.3599)	-.0215 (-.1884)	0.9916	376.69
Cote d'Ivoire	.3171 (.0922)	.0273 (.3900)	0.1922 (-.7392)	.9682 (18.530)	.0070 (.1040)	.1819 (.6329)	0.9808	163.09
Kenya	3.8985 (1.0384)	-.0432 (-.6249)	.0574 (.3850)	.9363 (15.698)	-.0680 (-1.0872)	-.0428 (-.2865)	0.9603	77.47
Madagascar	-5.5819 (-.2250)	.0043 (.1597)	.0303 (.1561)	.9730 (36.477)	.0042 (.1577)	-.0261 (-.1304)	0.9922	407.19
Malawi	2.6444 (1.1336)	-.0818 (-.8219)	-.2103 (-.7700)	.9311 (13.472)	-.1767 (-1.1931)	.2485 (.7113)	0.9635	84.37
Mauritius	5.8292 (.3125)	-.0078 (-.0766)	-.0912 (-.2741)	.9919 (10.547)	-.0276 (-.2276)	.0648 (.1800)	0.9109	32.70
Nigeria	.8941 (.1522)	.0320 (.3769)	-1.6198 (-3.0769)	.9627 (11.598)	-.0228 (-.2380)	1.7957 (3.0730)	0.9447	54.69
Sierra Leone	3.1237 (.7635)	-.0157 (-.3161)	-.0821 (-.5562)	.9183 (23.037)	-.0354 (-.7537)	.0931 (.5404)	0.9773	137.59
Swaziland	2.2988 (.2135)	-.0091 (-.1465)	-.0269 (-.2171)	.9941 (17.202)	-.0054 (-.1003)	.0027 (.0216)	0.9636	84.83
Zambia	-1.0163 (-.2874)	.0012 (.0242)	-.0691 (-7856)	.9376 (22.980)	-.0039 (-.0876)	.1227 (.9182)	0.9765	132.98
Zimbabwe	-3.4638 (-.3557)	.0085 (.2066)	.0646 (.3032)	1.0229 (24.725)	.0062 (.2125)	-.0545 (-.2671)	0.9795	152.92

Note: Figures in brackets beneath the estimated parameters refer to the t-ratios

Table 6. Hypotheses: __{2 =} __{2 =} __{2 = 0}

	_	_1	_o	_1	_o	_1	R2	F
Botswana	35.079 (2.7309)	.5343 (.5134)	.0209 (.0975)	-.5550 (-1.8260)	-.8504 (-.8599)	-.0064 (-.0273)	0.3300	1.58
Cote d'Ivoire	66.246 (6.9577)	-.8613 (-.8445)	-.3931 (-2.2530)	-.7066 (-5.4677)	.0406 (.0433)	-.4796 (-2.5492)	0.8578	19.30
Kenya	65.0911 (7.8471)	-.7205 (-1.3618)	-.4817 (-3.3832)	-.6937 (-4.8615)	-.0956 (-.2199)	-.5812 (-4.1432)	0.7787	11.26
Madagascar	1205.0 (13.780)	-.4036 (-.5639)	-.8581 (-9.8267)	-.9018 (-9.5808)	-.4986 (-.7287)	-.8866 (-9.8824)	0.9035	29.97
Malawi	29.293 (4.6089)	-1.0396 (-1.3979)	-.5033 (-2.1277)	-.8461 (-4.1585)	.2921