The Impact of Economic Sanctions on South African Exports

Literature › Works

Simon J. Evenett (2002). The Impact of Economic Sanctions on South African Exports. Scottish Journal of Political Economy, Vol. 49, No. 5. Blackwell Publishers (Scottish Economic Society).

Type	journal article
Scope	Economics and economic history of apartheid
Method	econometric
Period studied	1978–1999
Regions	South Africa, United States, European Community (France, Italy, Great Britain), Nordic countries (Denmark, Norway, Sweden), Japan
Themes	sanctions divestment, methods data sources, comparative theory, macro fiscal finance debt
Citation key	evenett2002

Argument

Using a gravity-equation model and a sample of eight industrialised importers and 31 middle-income exporters, Evenett estimates the effect of sanctions on South African exports through importer-exporter-specific dummies for 1978-1999. A naive full-sample estimate suggests European Community sanctions hurt South African exports most, but this result depends on outlier bilateral pairs. Trimming the sample by the highest and lowest values of the dependent variable shows that US sanctions (the Comprehensive Anti-Apartheid Act) had the largest and most robust effect, cutting bilateral imports by about a third, while Nordic and Japanese sanctions show little significant effect. The effects appear persistent after sanctions were lifted, especially for the US, which suggests hysteresis in trade relationships. The paper also shows that outliers inflate estimated distance elasticities in gravity models.

Evidence and method

Bilateral import data for 1978-1999 from the World Bank Trade and Production Database (Olarreaga and Nicita), originally UN COMTRADE; national income data from World Development Indicators; great-circle distances between capitals. Sample is 8 importers by 31 middle-income exporters (248 pairs per year). Sanctions chronology comes from Hufbauer, Schott and Elliott (2002).

Place in the debates

• Effectiveness of trade sanctions against apartheid South Africa (see Sanctions, disinvestment and the anti-apartheid movement). Finds that US sanctions had a large, robust and persistent negative effect on South African exports (about one third), while EC effects are not robust and Nordic and Japanese effects are insignificant. This is more favourable to sanctions' trade impact than Levy's 'trivial' thesis, though it does not address the effect on regime change.
• Whether trade sanctions create incentives for reform, and whether removal restores trade (see Sanctions, disinvestment and the anti-apartheid movement). Suggests trade flows need not return to pre-sanctions levels after sanctions are lifted (hysteresis), which weakens the incentive value of promised sanctions relief.
• Methodology: outliers and distance elasticities in gravity equations (see Data reliability and econometric method). Argues that outliers inflate estimated distance elasticities and sanction dummies, and that the elasticity falls by about half once outliers are excluded.

Findings

• Including outliers wrongly makes EC sanctions look the most damaging; the robust result is that US sanctions mattered most. (p. 1)

  In fact, robustness checks reveal that the United States’ Comprehensive Anti-Apartheid Act played the largest role, reducing bilateral imports by a third.

• US sanctions did not give way to recovery after removal in 1991, whereas EC sanctions may have been followed by a partial bounce back. (p. 2)

  The estimates also suggest that South African exports have not recovered after the removal of US sanctions in 1991. In contrast, the end of EC sanctions may have led to some ‘bounce back’

• Sanction-driven declines persist after removal, implying weaker incentive effects from the promise of lifting sanctions. (p. 2)

  the promise of the removal of sanctions if foreign behaviour improves may provide weaker incentives than previously thought, as trade flows need not return to their pre-sanctions levels.

• The EC result in the full sample collapses as outliers are trimmed: with c=0.1 only two of six sanction-year estimates are significant. (p. 11)

  When c = 01, however, we can only make a comparable claim for two out of six of these estimated annual elasticities. So much for the robustness of the previous findings on the EC sanctions regime.

• US sanctions parameter becomes increasingly significant over time once outliers are trimmed, with no significant negative estimates before sanctions but seven of eight afterwards. (p. 11)

  In 1978–1985, no annual sanctions parameter was both negative and statistically significant. During 1992–1999, seven out of eight were.

• Nordic and Japanese sanctions show no statistically significant effect on South African exports, and outliers are not responsible for this. (p. 14)

  outliers are not responsible for the finding that, by and large, the Nordic and Japanese regimes had no statistically significant effect on South African exports.

• With outliers, EC sanctions appear to halve South African exports; without them the effect falls to 27%, while US sanctions reduce exports by about a third. (p. 16)

  once outliers are excluded, the EC sanctions regime is found to reduce South African exports by 27%; still a sizeable impact but not a 50% cut. In contrast, US sanctions are found to reduce the apartheid regime’s exports the most; essentially by a third.

• Compared with other middle-income exporters, South African exports to the US fell further behind after sanctions were removed. (p. 16)

  compared to other middle income developing economies, South African exports to the US market have fallen further behind since the elimination of sanctions.

• Overall, only for the US is there considerable and fairly robust evidence of a statistically significant sanctions effect during 1986-1991. (p. 16)

  Overall, then, only in the case of the United States does considerable and fairly robust evidence point towards statistically significant effects of her sanctions regime on South African exports during 1986–1991.

• Outliers inflate the gravity distance elasticity, which falls by about half when 10% of pairs are trimmed at each tail. (p. 2)

  the distance elasticity falls 50% once the outliers are excluded, going some way to remedy the implausibly large size of typical gravity-inspired estimates of this elasticity

Evidence ledger

Sources used

Source	Kind	Period	How used
World Bank Trade and Production Database (bilateral imports, 1978-1999), original source UN COMTRADE	international trade or finance data	1978-1999	Supplies the dependent variable, bilateral import values, for the annual gravity-equation cross-sections used to estimate sanctions effects.
World Bank World Development Indicators (volume and CD-ROM): income classification, national income, US price deflator	international trade or finance data	1978-1999	Used to classify countries as middle income, to select the exporter sample, to supply national income data for the gravity equation, and to deflate values to 1995 US dollars.
Great Circle distance data between capitals (gravity-equation distance file, Haveman trade resources)	maps spatial data	time-invariant	Proxies bilateral trade barriers/transport costs through the distance term in the gravity regression.
Hufbauer, Schott and Elliott (2002), Case 62-2 and Case 85-1: Notes on sanctions against South Africa in the post-war era (mimeo, Institute for International Economics)	secondary literature compilation	1962-1993	Provides the chronology of sanctions (Table 1) that fixes the pre-sanction, sanction and post-sanction periods and which countries are treated as sanctioners.

New data

Author assembled annual bilateral trade matrices for 1978-1999 between 8 importing industrial countries and 31 middle-income exporters (248 bilateral pairs per year), combined with national incomes, distance and a sanctions-country-pair dummy structure. The trade data were supplied by Nicita from the World Bank database.

Inference

Quantification: Econometric (causal design).

Year-by-year cross-sectional gravity equations with importer and exporter dummies, with sanction effects captured by importer-South Africa pair dummies. Before/during/after comparison of the estimated dummy across 1978-85, 1986-91 and 1992-99, with outlier-trimming robustness checks at cutoffs of 1% to 10% of the ranked dependent variable. Effects are summarised as median coefficients and implied percentage export reductions. This is a difference-in-periods design with no formal control for other determinants.

Key claims

Claim	Basis	Strength
With the full sample, only EC sanctions parameters are negative and significant, suggesting the EC sanctions had the largest effect, a roughly 50% cut in South African exports. (p. 565)	Table 3 year-by-year gravity estimates on all 248 pairs, medians by period; Table 6 at cutoff 0.	Systematic data
The EC result is not robust to outlier exclusion; the number of negative, significant EC parameters falls sharply as the cutoff rises, and the implied effect drops to about 27%. (p. 567)	Table 4 re-estimation at c=0.01 to 0.10; Table 6 comparison of c=0 and c=0.1.	Systematic data
US sanctions (the Comprehensive Anti-Apartheid Act) had the largest effect, reducing South African exports by about a third, once outliers are removed. (p. 572)	Tables 4 and 6 at cutoff 0.1, where the US coefficients become more often negative and significant.	Systematic data
Nordic and Japanese sanctions had no statistically significant effect on South African exports. (p. 570)	Tables 3 and 4: sanction parameters are mostly insignificant across all cutoffs.	Systematic data
South African exports did not recover after US sanctions were lifted, while EC removal brought a small recovery, suggesting hysteresis in trade patterns. (p. 572)	Table 6 comparison of 1986-91 and 1992-99 implied reductions at c=0.1 (US 33.4% to 43.9%; EC 27.2% to 22.1%). The explanation via divestment and new suppliers is conjecture.	Partial data
Outliers inflate the estimated distance elasticity, which falls by half when the extreme observations are trimmed. (p. 570)	Table 5 median distance parameters across cutoffs, with the sample falling from 248 to 199 pairs.	Systematic data
Part of the pre-sanction reluctance to buy South African goods reflected private-sector behaviour, so the estimated sanction effect combines state sanctions with private boycotts and divestment. (p. 565)	Negative and significant EC dummies in 1978-85 in Table 3, plus theoretical argument. No direct data on firms' sourcing decisions or divestment.	Illustrative or anecdotal

The author's caveats

The author notes that the dummy is only an imperfect proxy for sanctions, that distance imperfectly proxies transportation costs, and that price-index dummies should not be read literally. He also notes that the Nordic, Japanese and US sanction parameters are imprecisely estimated in the full-sample runs. Zero-trade observations may create the outliers. The distance-outlier results are not reported ('Considerations of space') but are offered on request.

Gaps in the evidence

The study cannot separate state sanctions from private boycotts, divestment, capital controls, political instability or South African domestic policy, because it relies on a single pair dummy. Disaggregated trade data by product (sanctioned versus unsanctioned goods such as gold, coal, iron and steel), which exist in COMTRADE, would test the mechanism directly. The outlier trimming is ad hoc and is no substitute for modern methods for zero trade (e.g. PPML) or for panel specifications with pair fixed effects. Sanctions by other countries (Canada, Australia, Germany) and diversion of exports to third markets are not analysed.

Replicability: Data described but not public.