8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 1/12
tradeliberalisation
DO frEE TrADE AGrEEMENTS
CREATE TRADE FOR SOUTH AFRICA?
B Y M E R L E H O L D E N A N D L A N D O N M C M I L L A N
U N I v E R S I T Y O F k wA z U L U - N AT A L D U R B A N & T I P S
1. Introduction
wHILE TRADE POLICY IN SOUTH AFRICAhas been determined in a multilat-eral setting under the auspices o the Uruguay Round and the WTO, pre -erential trading agreements have also been pursued as evidenced by theSADC Free Trade Agreement and the EU SA FTA. In addition, South A rican
e porters have recently benefted rom pre erential access into the US under the A rican Growth
and Opportunity Act (AGOA). These agreements are relatively recent events, or South A rica is alsoa member o the longest standing customs union in the world – the Southern A rican CustomsUnion (SACU), consisting o South A rica, Botswana, Lesotho, Swaziland and Namibia. However,despite multilateral commitments, policy-makers in South A rica continue to negotiate pre erentialagreements with countries as diverse as the US, India, Brazil and more recently China. Whetherthese agreements will act as a substitute or MFN trade policy remains to be seen but many econo-
mists view these developments with a degree o scepticism given their potential or trade diversion
as well as trade creation.
The analysis in this paper supports the
that the EU-SA FTA stimulated both e po
imports, while or SADC e ports were sti
but the results or imports was ambiguou
AGOA results were ar less signifcant
suggesting that pre erential access or
A rican e ports into the US had not
particularly benefcial.
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 2/12
Trade&IndustryMonitor
Tr d Lib r li tion
This paper attempts to ascertain whether the two ree trade agreements,
the EU SA FTA, and SADC have had any signifcant impact on trading
patterns or South A rica. Furthermore, although the AGOA arrangements
with the United States do not take the orm o a ree trade agreement, by
estimating the impacts on e ports and imports separately it is possible to
actor this agreement as well into the analysis..
Although these agreements all came into orce in 2000/2001 a ter a peri-
od o phased in aggressive trade liberalization under the Uruguay Round,
the view may be that insu fcient time had passed to enable the impact
to be elt. Nevertheless it is our view that su fcient time has elapsed or
e porters and importers to have responded and given that the EU-SA
agreement is due or review, assessment o these agreements will add
to the debate.
There ore the paper ocuses on whether e ports rom South A rica have
increased and whether imports have similarly been a ected by the grant-ing o pre erential access under the agreements. In addition, the research
is able to estimate the role played by changing transport costs over the
period 1994 to 2004.
E ports Imports
50,000
45,000
40,000
5,000
0,000
24,000
20,000
15,000
10,000
5,000
- 1994 1995 1996 1997 1998 1999 2000 2001 2002 200 2004
M i l l i o n s
Figure 1 : TOTAL ExPORTS AND IMPORTS FOR SOUTH AFRICA: 1994 - 2004 (US$-MILLION)
2. Trends in trade
The ollowing graphs provide a picture o the trends in South A rica
ports and imports rom 1994 to 2004.
Figure 1 demonstrates that both imports and e ports or South A rica eperienced signifcant growth in the years 200 and 2004.
Figure 2 shows that the EU and rest o the world e ports increased mark
edly in 200 and 2004 despite the strength o the rand during this pe-
riod.
Figure shows an increase in imports predominantly rom the EU (esp
cially Germany) and the rest o the world (especially China, see Appen
1 Figure A1). While imports rom Germany demonstrate signifcant grow
over the period, growth o imports rom China has been greater. Whe
the growth o e ports and imports rom China are compared, clearly th
balance o trade is turning in the Chinese direction providing evidence
increasing Chinese investment in the South A rican economy.
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 3/12
Do fr Tr d agr m nt Cr t Tr d or south a ric ?
Note: Rest o the world imports only include those countries in the dataset.
25,000
20,000
15,000
10,000
5,000
0 1994 1995 1996 1997 1998 1999 2000 2001 2002 200 2004
sum o sadcIMP sum o euIMPsum o USIMP sum o RestIMP
M i l l i o n U
S $
Figure 3 : IMPORTS BY REGION: 1994 - 2004 (CURRENT US$-MILLION)
Breakdown o South A rican imports by region(US$-million)
20,000
15,000
10,000
5,000
0
1994 1995 1996 1997 1998 1999 2000 2001 2002 200 2004
sum o sadcExP sum o euExPsum o agoaExP sum o RestExP
M i l l i o n U S $
Breakdown o South A rican e ports by region(US$-million)
Figure 2 : ExPORTS BY REGION:1994 - 2004 (CURRENT US$-MILLION)
Note: Rest o the world e ports only include those countries in the dataset.
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 4/12
4
Trade&IndustryMonitor
Tra Lib ralisati n
(1)
For South A rican imports M rom country i at time t
(2)
where
EU, SADC and AGOA are binary dummy variables which are unity or member countries (aside rom South A rica) o the two groups and unity or South A
to the US
Ys are sets o variables rom gravity models
and are well behaved residuals.
Feenstra (2002) divides estimation o a gravity type equation in three approaches. Firstly, price indices can be used to measure price e ects (Bergstrand, 1989); secondly bord
can be used to measure price e ects indirectly (Anderson and van Wincoop, 2002), and fnally, fxed e ects can be used or the source o imports and destination o exports.
di fculties o specifcally modeling the price implications o the agreements, the second approach was used. We ruled out the use o fxed e ects or reason o wishing to es
impact o distance on South A rican trade.
More recently the division o markets spatially has been success ully modeled by applying gravity variables relating to economic mass and we ollow this approach in the p
variables included are as ollows: relative GDP, population, and the distance between trading partners. (Rose 2004, Anderson and Wincoop, 2002). Using border e ects or th
agreements in the orm o dummy variables, once the agreements were e ective, it is possible to ascertain whether the trading agreements had shi ted trade or South A ric
Thus, the conditioning variables Y above include the specifc country GDP vis a vis South A rican GDP; specifc country population vis a vis South A rican population and the dis
the South A rican port to the specifc country port o entry. When the estimating equation is expressed in log linear terms it ollows that
(�)(�)
For the export equation the AGOA dummy is included whereas or imports it is excluded. The South A rican data is represented by the subscript j and the trading partner
subscript i.
MeThodoLogy
Given that South A rica has recently entered into two ree trade agreements, one with the South A rican Development Community (SADC) and the other with the European U
a methodology that attempts to establish the trade increasing e ects o both o these agreements relies primarily on application o several regression models.
The regressions distinguish the impact o the agreements based on panel data or South A rican exports and imports over the period 1994 to 2004. The evidence is derived rotions o the ollowing orm:
For South A rican exports X to country i at time t
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 5/12
Do fr Tr d agr m nt Cr t Tr d or south a ric ?
3. Estimation results
The data are summarised in Table 1. Given that suitable de ators are not
available or trade rom and to the trading partner countries, South A rican
e ports and imports are measured in nominal US dollars converted at av-
erage or the year market e change rates. GDPs are measured in constant
US dollars and distance in thousands o kilometres. Data sources and the
list o countries can be ound in Appendi 1.
1994 1996 1998 2000 2002 2004
ex ort (Us$-million)
Mean 1 1.40 158.28 15 .64 186.21 159. 9 288.10
Standard deviation 27.69 78.24 87.04 500.41 82.67 7 2.54
Minimum 0.00 0.00 0.00 0.00 0.11 0.04
Ma imum 1 857.06 2 976.76 2 721.74 605. 8 2 411. 8 4 529. 9
Im ort (Us$-million)
Mean 156.0 196.9 189.88 196.55 189.09 45.01
Standard Deviation 497.92 591.60 54 .12 541.94 5 6.09 910.26
Minimum 0.00 0.00 0.00 0.00 0.00 0.00
Ma imum 7 4.16 4 050.91 752.04 585.16 4 066.71 6 728.66
GDp (Us$-million)
Mean 188 996 200 848 212 875 228 46 2 5 412 249 56
Standard deviation 786 056 8 419 891 8 9 956 5 9 979 01 1 0 7 948
Minimum 86 85 99 108 108 127
Ma imum 7 775 500 8 271 400 9 012 500 9 764 800 10 42 992 10 66 625
po ul tion (million )
Mean 8.68 9.78 40.84 41.87 42.72 4 .85
Standard Deviation 1 1.61 1 5.15 1 8.62 141.88 144.92 148.51
Minimum 0.01 0.01 0.01 0.01 0.01 0.01Ma imum 1 191.84 1 217.55 1 241.94 1 262.65 1 280.40 1 0 .55
Di t nc (km’000)
Mean 11.12
Standard Deviation .6
Minimum 1.50
Ma imum 19.00
Table 1: SUMMARY STATISTICS OF THE MODEL DATA 1994 - 2004 FOR 1 6 COUNTRIES
3.1 Cross-section results
Initially the model in equation was estimated as individual cross sec
tions or each year or both e ports and imports. The results are show
in Tables 2 and . Table 2 shows the estimation results or e ports ro
South A rica to 1 6 countries or each year rom 1994 to 2004.
The estimated coe fcients on relative GDPs and population were o th
e pected signs rom gravity equation theory, namely that GDP is signi
cantly and positively related to trade while populations were negatively
related. While larger countries tend to be more sel su fcient, the emer
gence o countries such as China and India as major trading nations is
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 6/12
Trade&IndustryMonitor
Tr d Lib r li tion
altering the negative signifcance o population. Distance is also shown
to be signifcant and, interestingly, surprisingly high when compared with
other estimates or gravity equations.
The elasticity on the distance variable or e ports is estimated to decline
rom - 4.9 in 1994 to - .11 in 2004. Estimates o the distance elasticity
in other studies is usually in the range o -0.5 to -1.0. (Coe et al, 2002).
The high elasticity or South A rican e ports demonstrates the remoteness
o South A rica located as it is at the tip o A rica, rom its major markets
in Europe, the US and East Asia. It is encouraging however that contrary
to other studies (See Coe et al, 2002 ) that have estimated constant dis-
tance coe fcients over time, that the South A rican elasticity or e ports
is declining.
More importantly the coe fcients on the dummy variables or the ree
trade agreements appear to demonstrate that in only one year was pre er-
ential access important or South A rican e ports. EU e ports were shownto be a ected by the ree trade agreement in 2002. On the other hand
neither the SADC nor AGOA access were signifcant demonstrating in the
frst place that access into the US market had not been particularly benef-
cial given the caveats attached to such pre erential access. In the case o
SADC this result is not surprising given the asymmetrical phase in o the
agreement. Once the estimations were done in panel orm, these results
GDPs Populations Distance Constant EU SADC AGOA
ex ort d nd nt v ri bl
1994 1.7 * -0.54* -4.9 * -21.91* - - -
1995 1.52* -0. 4*** -4. 4* -19.25* - - -
1996 1.46* -0. 4** -4.29* -17.75* - - -
1997 1.40* -0. 5** -4.09* -16.68* - - -
1998 1.48* -0.4 * -4.29* -17. 5* - - -
1999 1. 1* -0.24*** - .85* -15.51* - - -
2000 1. 9* -0. 4* -4.14* -16.01* - - -
2001 1.28* -0.21*** - .90* -14.75* 0.26 0. 4 -0.15
2002 1.08* -0.18** -2.96* -12. 4* 0. 8 0.96** 0.40
200 1.09* -0.12 - .25* -12.52* 0.50 0.64 0.
2004 1.11* -0.15 - .11* -1 .16* -.09 0.64 0.25
Note:
(1) * Signifcance at 1% level or higher
(2) ** Signifcance at 5% level
( ) *** Signifcance at 10% level
(4) Robust standard errors computed
Table 2: CROSS-SECTION ESTIMATES FOR ExPORTS: 1994 - 2004
were ound to be more avourable towards the impact o the agreemen
on trade over the entire period.
Table shows the estimations or South A rican imports. Again the G
and population variables behave as e pected rom gravity equations. Th
distance variable is also a signifcant determinant o imports but inter-
estingly the elasticity on the variable is not as high as it is or e port
Although the elasticity declines rom - .66 in 1994 to -2.7 in 2004 thi
decline is not continuous. We are o the view that the di erence in th
distance elasticities between e ports and imports re ects the di eren
in the composition o trade. South A rica is a major e porter o bulk lo
value goods while importing higher value intermediates, consumer and
capital goods hence accounting or the di erence in these coe fcien
The higher elasticity o the distance coe fcient or e ports also re e
the impact o competition rom suppliers o substitutable goods that a
closer to the main markets o the developed world. Whereas South A ri
sources most o its imports rom the developed world, and there ore h
less choice in substituting sources o supply that are closer.
The EU dummy in the import equation is ound to be signifcant in the
years 200 and 2004, probably due to the asymmetrical phase-in that
avoured South A rican access into the EU and delayed EU access in
the South A rican market. There ore, the fnding o a later response on t
import side is not surprising.
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 7/12
Do fr Tr d agr m nt Cr t Tr d or south a ric ?
GDPs Populations Distance Constant EU SADC
Im ort d nd nt v ri bl
1994 1.85* -0.60* - .66* -27.77* - -
1995 1.66* -0.42* -2.89* -25.82* - -
1996 1.65* -0.44* - .0 * -24.88* - -1997 1.66* -0.44* - .04* -24.74* - -
1998 1.76* -0.54* -2.58* -27.29* - -
1999 1.7 * -0.55* -2.89* -25.87* - -
2000 1.81* -0.54* -2.7 * -27.95* - -
2001 1.81* -0.58* - .0 * -26.90* 0.09 0.28
2002 1.58* -0. 5** -2.60* -2 .61* 0.18 0.2
200 1.6 * -0. 4* -2.71* -25.99* 0.70** 1.07
2004 1.66* -0.26** -2.7 * -27.05* 0.72** 1.09
Table 3: CROSS-SECTION ESTIMATES FOR IMPORTS: 1994 - 2004
Note:
(1) * Signifcance at 1% level or higher
(2) ** Signifcance at 5% level ( ) *** Signifcance at 10% level
(4) Robust standard errors computed
3.2 Panel estimations
Given the nature o the data, it was also decided to run panel estimations
o the gravity equations or both e ports and imports. The distance vari-
able is time invariant, there ore f ed e ects estimations were not used as
we were interested in estimating the impact o distance in the model and
checking this against the cross section work1. Generalised least squares
estimations assuming heteroskedastic panels with common autoregres-
sive frst order coe fcients o all panels were per ormed on the data
(Tables 4 and 5). The estimated coe fcients o the gravity variables were
ound to be similar to the average o the annual estimates suggesting that
the pooling o the data into the panels is valid.
In Table 4, since the regressand is the natural logarithm o e ports, the
impact o the agreements over the period since the agreement had
been in place is estimated to raise South A rican e ports by per-
cent or the EU (since e p (0.286) -1 = 0. ), by 50 percent or SADC(since e p (0.405)-1 = 0.50) and percent or the AGOA2
(since e p (0.0 2)-1 = 0.0 ). The elasticities on the other variables are o
the e pected sign and signifcance. The elasticity on the distance variable
in the panel estimates is - . or e ports.
In fxed e ects estimations time invariant variables are dropped. 2 The coe fcient on the AGOA variable was ound not to be signifcantly di erent rom zero
anyway.
In Table 5, the impact o the agreements over the period is estimated to
raise South A rican imports by 5 percent (since e p (0.42) -1 = 0.5
rom the EU and 80 percent rom SADC (since e p (0.59) -1 = 80). B
the ree trade agreements had stimulated trade or those countries sellin
into the South A rican market.
Once again the elasticities on the other variables were ound to be o th
e pected sign and signifcance. The distance elasticity or imports is -2.4
This value is lower than in the cross section work and certainly lower tha
it is or e ports.
While our data or e ports and imports are measured in nominal term
namely US dollars, we included time e ects to absorb the e ects o i
tion in the ollowing regression. (See Frankel et al, 1997 and Coe and
Ho maister, 1999). Thus, additional linear time trends were included
the distance, GDP and population variables as well as the dummy vari-
ables. These results are shown in Tables 6 and 7 or e ports and imporrespectively. Once again the estimated coe fcients are similar to the aver
age annual estimates in the cross section estimates suggesting that the
time dummies do address the problem o de ating the trade data.
The addition o the linear time trends shows that, where they are signif-
cantly di erent rom zero, the signs o the interactive time trend variab
imply that the absolute value o the estimated coe fcients decline over tim
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 8/12
Trade&IndustryMonitor
Tr d Lib r li tion
The distance variable is initially signifcantly negative and becomes less so
over time (because the trend is positive). The SADC, EU and AGOA, in con-
trast, are signifcantly positive and declining over time. This fnding is onlyconfrmed in the cross section estimations or the distance variable.
The inclusion o the interactive linear time trends into the import equa-
tion led to much the same results as previous estimations, e cept or the
SADC dummy and its time trend and the distance time trend. The SADC
time trend appears to have completely reduced the value o the access or
SADC members into the South A rican market.
This fnding could be related to the view that many SADC countries had, prior
to the agreement, enjoyed some pre erential access under bilateral agree-
ments with South A rica and there ore the value o additional access under
the FTA was low4. The EU agreement remains a signifcant determinant o
3 The distance, EU, SADC and AGOA trend variables are all ound to be signifcantly di erent rom zero. The act that the trend variables on GDP and population were insig-nifcant suggests that their explanatory e ects remained relatively constant over time.
4 Zimbabwe, Zambia, Malawi and Mozambique had agreements that had been negotiated during the apartheid era.
Variable Coe fcient Standard Error z P>z
Log GDPs 1.08 0.0 6.97 0.000
Log populations -0.06 0.0 -2.07 0.0 9
Log distance - . 0.11 - 1. 2 0.000
EU 0.286 0.07 .80 0.000SADC 0.405 0.115 .54 0.000
AGOA 0.0 2 0.22 0.15 0.88
Constant -12.60 0.44 -28.77 0.000
Table 4: PANEL ESTIMATES FOR LOG ExPORTS AS DEPENDENT VARIABLE
Note: Coe fcients were estimated by generalised least squares assuming heteroskedastic panels with common autoregressive frst-order coe fcients o all panels.
Variable Coe fcient Standard Error z P>z
Log GDPs 1.46 0.02 50.72 0.000
Log populations -0.29 0.0 -9.21 0.000
Log distance -2.41 0.097 -24.65 0.000
EU 0.42 0.071 5.86 0.000
SADC 0.59 0.15 .89 0.000
Constant -22.41 0.47 -47.59 0.000
Table 5: LOG IMPORTS AS DEPENDENT VARIABLE
Note: Coe fcients were estimated by generalised least squares assuming heteroskedastic panels with common autoregressive frst-order coe fcients o all panels.
imports into South A rica. Interestingly, the impact o distance on impo
over time, in contrast to e ports is ound to be insignifcantly di erent
zero.
On closer e amination o the cross-section estimates, the distance co
e fcients are not monotonically declining over the period. The act tha
the population trend is more signifcant and positive or imports than o
e ports is a re ection o the increased proportion o imports originati
in China.
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 9/12
Do fr Tr d agr m nt Cr t Tr d or south a ric ?
Coe fcient Standard Error Z P> Z
Log GDPs 1. 0 0.02 60.27 0.000
Log GDPs trend -0.002 0.00 -0.57 0.570
Log population -0.26 0.0 1 -8. 6 0.000
Log populations trend 0.005 0.005 1.04 0. 00
Log distance -4.29 0.1 2 - 2. 9 0.000
Log distance trend 0.06 0.022 2.64 0.008
EU 1.54 0. 75 4.09 0.000
EU trend -0.18 0.04 -4.4 0.000
SADC 2. 7 0.89 2.66 0.008
SADC trend -0.199 0.09 -2.11 0.0 5
AGOA 1.01 0.52 1.95 0.052
AGOA trend -0.15 0.05 -2.74 0.006
Constant -15.04 0.19 -79.27 0.000
Table 6: LOG OF ExPORTS AS DEPENDENT VARIABLE
Note: Generalised least squares estimates o coe fcients with heteroskedastic panels. The trend is equal to 1 in 1994, 2 in 1995 … and 11 in 2004.
Coe fcient Standard Error Z P> Z
Log GDPs 1.69 0.022 75.27 0.000
Log GDPs trend -0.00 0.00 -0.19 0.851
Log population -0.54 0. 4 -15.84 0.000
Log populations trend 0.01 0.005 2.14 0.0 2
Log distance -2.92 0.102 -28.80 0.000
Log distance trend 0.02 0.019 1. 4 0.18
EU 1.77 0.466 .79 0.000EU trend -0.17 0.049 - .49 0.000
SADC 0.187 0.912 0.21 0.8 7
SADC trend 0.06 0.098 0.57 0.57
Constant -25.70 0.288 -89.28 0.000
Table 7: LOG OF IMPORTS AS DEPENDENT VARIABLE
Note: Generalised least squares estimates o coe fcients with heteroskedastic panels. The trend is equal to 1 in 1994, 2 in 1995 … and 11 in 2004.
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 10/12
8/6/2019 Evaluation of Benefits of Trade Agreements for S.a.
http://slidepdf.com/reader/full/evaluation-of-benefits-of-trade-agreements-for-sa 11/12
Do fr Tr d agr m nt Cr t Tr d or south a ric ?
Appendix 1: Countries and data sources
Albania Ecuador Malaysia Thailand
Algeria Egypt Maldives Togo
Angola El Salvador Mali Trinidad and Tobago
Anguilla Estonia Malta Tunisia
Antigua and Barbuda Ethiopia Mauritius Turkey
Argentina Finland Me ico Uganda
Armenia France Moldova UkraineAustralia Gabon Morocco United Arab Emirates
Austria Germany Mozambique United Kingdom
Bahamas Ghana Nepal United States
Bangladesh Greece Netherlands Uruguay
Barbados Guatemala New Zealand Venezuela
Belgium Guinea Niger Vietnam
Belize Guinea-Bissau Nigeria Yemen
Benin Guyana Norway Zambia
Bolivia Haiti Oman Zimbabwe
Brazil Honduras Pakistan
British Virgin Islands Hong Kong PeruBulgaria Hungary Philippines
Burkina Faso Iceland Poland
Burundi India Portugal
Cambodia Indonesia Republic o Korea
Cameroon Iran Romania
Canada Ireland Russian Federation
Cayman Islands Israel Samoa
Central A rican Republic Italy Saudi Arabia
Chile Jamaica Taiwan Province o China
China Japan Tanzania
Colombia Kazakhstan Senegal
Comoros Kenya Seychelles
The countries included in the study numbered 1 6 in total. These coun-
tries had traded with South A rica over the period 1994 to 2004. During
some years when either imports or e ports were zero log values were
taken on the value one.
GDP, population and South A rican e ports and imports by trading partn
were obtained rom the TIPS database. Distance between trading partner
and South A rica were obtained rom Holden (1996).