Body
Economic sanctions have resurfaced at the center of public policy debate. After a brief lull following the politically disastrous grain embargo and pipeline sanctions in the early 1980s, sanctions are once again the weapon of choice to enforce a myriad of US foreign policy goals, from countering terrorism to battling drug trafficking. A recent National Association of Manufacturers (1997) study lists over 30 countries hit by new US sanctions during the period 1993-1996. Many of these actions were unilateral, reducing their impact in an increasingly globalized economy that has many alternative suppliers and markets. High-publicity initiatives, such as the Helms-Burton Act and the Iran/Libya Sanctions Act, which threaten to punish third-country corporations that conduct business in Cuba, Iran, and Libya, also raise the possibility that frustrated OECD governments (such as Canada and France) will retaliate against US companies.
In an increasingly integrated global economy, it is important to have a clear understanding of the costs and benefits of unilateral economic sanctions for the United States. Most of the analysis of the effectiveness of economic sanctions suggests they have limited utility for changing the behavior or governments of target countries. Previous research at the Institute for International Economics concluded that US sanctions had positive outcomes in fewer than one in five cases in the 1970s and 1980s.1 Much less is known about the costs of economic sanctions for the US economy. For that reason, the present analysis focuses on the cost side of the equation.2
This study aims to empirically measure the impact of economic sanctions on bilateral trade flows. The intent of trade sanctions is of course to reduce trade--exports or imports or both. Financial sanctions may also reduce trade by denying investment, foreign exchange or credit to the target country or by raising its cost of credit.
In addition to the immediate impact of sanctions on trade with the target, many American businessmen claim that the effects of even limited unilateral US sanctions go well beyond targeted sectors. They also argue that the effects linger long after they are lifted because US firms come to be regarded as "unreliable suppliers." Sanctioned countries may avoid buying from US exporters even when sanctions are not in place, thus giving firms in other countries a competitive advantage in those markets.
Exports lost today may mean lower exports after sanctions are lifted because US firms will not be able to supply replacement parts or related technologies. Foreign firms may also design US intermediate goods and technology out of their final products for fear of one day being caught up in a US sanction episode. The secondary boycotts and extraterritorial sanctions passed last year in the Iran/Libya Sanctions Act and the Helms-Burton Act targeting Cuba are disturbing precedents that could increase the unreliable supplier effect in the future. These indirect effects may well extend beyond the sanctioned products and even beyond the time period sanctions are imposed.
A common method in economics for analyzing bilateral trade flows (exports plus imports and exports alone) is the so-called "gravity model." Applying an even more common statistical technique, "ordinary least squares" regression analysis, to the gravity model allows us to isolate the effect of sanctions on bilateral trade flows while holding other factors constant, such as size and distance. The gravity model and the OLS technique permit us to analyze the indirect as well as direct effects of economic sanctions across a large number of countries.
This approach should be contrasted with the case study approach to estimating the trade effects of economic sanctions. A case study approach calculates trade interruptions that are identified by competent observers--for example, affected firms or responsible government agencies. The case study approach best captures anecdotal eye-witness reports but it may miss less visible secondary effects. In addition, it is difficult to reach general conclusions from a handful of cases.
To preview the results, we find that US sanctions in 1995 may have reduced US exports to 26 target countries by as much as $15 billion to $19 billion. If there was no offsetting increase in exports to other markets, that would mean a reduction of more than 200,000 jobs in the relatively higher-wage export sector and a consequent loss of nearly $1 billion annually in export sector wage premiums. This suggests a relatively high cost to the US economy while sanctions are in place.
However, we find only limited evidence that the negative impact of sanctions lingers long after they are lifted. This may reflect the highly aggregated nature of the data we use. Long-term effects of sanctions might be expected to be relatively more severe for particular sectors, such as sophisticated equipment and infrastructure, than for exports in the aggregate. And, as noted, continued use of extraterritorial sanctions could increase the effect for these sectors in the future. We also find, not surprisingly, that foreign firms have replaced US firms in Cuba and that Canada, Australia, and Germany export more to China than size, income, and geography would suggest.
The Gravity Model
With foundations in the physical sciences, the gravity model has consistently proved to be a useful tool for the analysis of bilateral trade flows. Isaac Newton originally devised the model to explain gravitational force in the universe. His theory states that the gravitational pull between two celestial bodies is positively related to the product of their masses and inversely related to their distance apart. Similarly, in its simplest form, the gravity model as applied to trade predicts that the amount of trade between two countries will be positively related to the product of their outputs (a measure of size or mass), and negatively related to the distance between them. The gravity model has been applied to bilateral trade since at least the 1960s and it has enjoyed a resurgence in the 1990s as an empirical tool for analyzing regional trading areas.
Most applications of the gravity model include other variables besides size and distance that might be expected to influence trade flows. The model used in this study predicts that bilateral trade will increase as combined size and per capita incomes increase, decrease as the distance between two countries increases, and increase if the two countries share a common border or a common language, or are both members of the same trade bloc (for example, the North American Free Trade Area or the European Union).3 In addition, we have added a series of dummy variables to capture the effect of trade sanctions. The following sections specifically define the variables and briefly describe their hypothesized effect on trade flows.
Dependent variable
The dependent variable in this model--the variable to be explained-- is bilateral merchandise trade (Trade), which is defined as exports plus imports expressed in current dollars, in each of three separate years: 1985, 1990, and 1995. Exports alone are available for 1995 and we also use those data as the dependent variable in some tests. Our data set includes 88 countries (listed in Appendix 1), which results in 3,827 different country pairs. For several statistical tests, selected subsets of the 3,827 country pairs were examined. The source for trade information is the International Monetary Fund's Direction of Trade Statistics. GNP and population figures are primarily taken from the International Monetary Fund's International Financial Statistics, supplemented when necessary by data from the World Bank's World Development Report and the CIA World Factbook.
Logarithm-linear form
The regression equation form is logarithmic-linear, or log-linear, meaning that the equation is linear when all variables are expressed either as logarithms or as dummy variables. The continuous variables discussed below (Trade, GNP, GNP per capita, and distance) are all expressed in logarithmic form.4 The regression coefficient on a continuous logarithmic variable can be interpreted as an elasticity, that is, as the ratio of the percentage change in the dependent variable for each one percent change in the independent variable. For example, if the estimated coefficient on log (GNP) is 0.9, a 10 percent increase in a country's GNP increases the dependent variable (merchandise trade) by 9 percent.
The dummy variables discussed below (e.g., for a common border or language, or for the presence of sanctions) take the value 1 or 0. Because the dependent variable -- Trade -- is expressed in logarithmic form, one must take the exponent of the coefficient of a dummy variable before interpreting it. The coefficient on a dummy variable can then be interpreted as a percentage shift in the dependent variable when the dummy takes the value 1. For example, if the coefficient on a dummy is 0.504, then, when the dummy takes the value 1, the dependent variable is 65.5 percent larger than otherwise (the base e to the exponent .504 = 1.655)
Core gravity model variables
- GNP is the country's gross national product, GNP, measured in dollars. Two large countries, all else equal, are expected to have a higher volume of trade than two small countries.
- GNPPC is the per capita GNP of the country, again measured in dollars, and defined as GNP divided by population. Trade tends to rise more than proportionally to GNP as a country becomes richer, and less than proportionally to GNP if the driving force behind a larger economy is simply an increase in population. One reason is that, as per capita income rises, individuals want to consume a wider variety of goods and services, which increases the demand for differentiated products. Wealthier countries also tend to have lower trade barriers than poorer ones, another reason why higher incomes and higher trade levels go together.
- DIST is the distance between the two countries, in most cases measured as the number of kilometers between the two capitals. Occasionally the capital is replaced by another major city that more closely approximates the geographical center for commercial activity (in the US case, Chicago). Greater distance tends to decrease trade, as transport costs and convenience favor closer sources and markets.5
- ADJ (adjacency) is a dummy variable equal to 1 if the two countries in a pair share a common border and 0 otherwise. A shared border facilitates trade to an even greater degree than simple nearness.
- LANG is a dummy variable equal to 1 if the two countries in a pair share a common language and 0 otherwise. A shared language facilitates commercial transactions and travel.
- BLOC is a dummy variable equal to 1 if the two countries in a pair belong to the same trading bloc (such as NAFTA or the EU) and 0 otherwise.
Basic regression equation
We used an ordinary least-squares (OLS) regression for this study. The main advantage of OLS analysis is that it can be used to estimate the independent effect of each factor, holding constant the effects of the other variables in the equation. In the following regression equation for the basic gravity model, the coefficients for the independent variables described above are represented by ß1 through ß6 , C signifies the constant, and the dependent variable is the logarithm of trade (exports plus imports, or exports only) between the two countries, country i and country j. Because of the logarithmic form of the equation, the coefficients on the continuous variables can be interpreted as elasticities, and the coefficients on the dummy variables can be interpreted as percentage shifts.
ß4 (ADJ) + ß5 (LANG) + ß6 (BLOC)
Using the log-linear form of the regression equation requires that the observations where bilateral trade is reported as zero must be dropped, because it is not possible to take the logarithm of zero. Trade might be recorded as zero because two small, distant countries actually do not trade, or because neither party reports trade to the international institutions that collect data. Eliminating these zero values reduces the number of available observations each year by at least 1,000. This might be expected to distort the regression results since some of the zero observations being dropped are likely to show no trade because of sanctions. Using an alternative method, we tested for the severity of this potential problem, and found it was not an important difficulty.6
Incorporating Sanctions
Finally, in order to test for the impact of economic sanctions on bilateral trade, we developed a set of nine dummy variables to indicate current or previous economic sanctions between the two countries. (Table 1 summarizes the variable definitions. Table 8 lists the sanctions episodes used to calculate the reduction in US exports.) Because sanctions take such a wide variety of forms, we first divided the cases into three categories according to the intensity or coverage of the sanctions.
- LIM, MOD, EXT indicate whether sanctions were in place during the year in question (1985, 1990, or 1995), and whether their coverage was "limited", "moderate," or "extensive." We considered minor financial, export, cultural, or travel sanctions to be "limited". Examples include suspending or reducing bilateral aid, and imposing export restrictions on weapons or narrow categories of dual-use technologies. Broader trade or financial sanctions were classified as "moderate". We generally reserved the "extensive" category for comprehensive trade and financial sanctions such as those against Iraq or Serbia. Sometimes, however, a combination of several "moderate" sanctions, such as US export controls against the Soviet Union and Eastern Europe during the Cold War, together with denial of MFN status under the Jackson-Vanik amendment, were together considered "extensive." The dummy variables do not distinguish between target and sender. They simply indicate whether sanctions exist between the pair of countries, and their severity. If both countries had sanctions in place simultaneously against each other, the dummy variable reflects the most severe level.
- LIM1*2, MOD1*2, EXT1*2 are used to represent cases where sanctions were not present during the year under analysis (1985, 1990, or 1995) but had been in place in the previous 1 to 2 years. The purpose is to discover whether sanctions continue to dampen trade even after they are removed. The same categories, "limited", "moderate" and "extensive", describe the degree of severity and coverage of the sanctions previously in place.
- LIM3*4, MOD3*4, EXT3*4 represent cases where sanctions had been lifted 3 to 4 years before 1985, 1990, or 1995.
With the sanctions dummy variables included, the regression equation becomes:
ß5 (LANG) + ß6 (BLOC) + ß7 (LIM) + ß8 (MOD) + ß9 (EXT) + ß10 (LIM1*2) +
ß11 (MOD1*2) + ß12 (EXT1*2) + ß13 (LIM3*4) + ß14 (MOD3*4) + ß15 (EXT3*4)
General Results
As in Frankel (1997), the basic gravity model variables--size, income, distance, adjacency, and language--all have the expected sign and are all highly significant statistically (at the 99 percent confidence level or better). The trade bloc dummy is highly significant in 1985 and 1990, but only marginally so in 1995. The equation as specified, including the sanctions dummies, explains 70 percent or more of the variation in observed bilateral trade flows (see Tables 2, 3, and 4).
Of primary interest here is the impact of economic sanctions on bilateral trade flows. As expected, when they are in place, extensive sanctions have a large impact on bilateral trade flows, consistently reducing them by around 90 percent. There is more variance in the estimated impact of moderate and limited sanctions and the results are not quite as robust, but they suggest an average reduction in bilateral trade of roughly a quarter to a third (Table 5).7 The coefficients on the dummy variables representing the presence of sanctions in the base years, 1985, 1990, and 1995, are highly significant statistically (at the 99 percent confidence level or better) with just two exceptions, moderate sanctions in 1990, which are still significant at the 95 percent level, and limited sanctions in 1995, which are statistically significant just below the 90 percent level.
There is only limited evidence, however, that sanctions continue to depress trade after they have been lifted. The results suggest that extensive sanctions lifted three or four years earlier (EXT3*4) reduced 1985 bilateral trade between the previous target and sender countries by nearly 90 percent. Unfortunately, there are no observations for extensive sanctions lagged three or four years in the 1990 and 1995 data sets, so there is no way to know if the 1985 result is anomalous. The coefficients for sanctions lifted one to two years previously (LIM1*2, MOD1*2, EXT1*2) generally have negative signs, as expected, but they are not statistically significant at the usual levels.8 Interestingly, there is some evidence of a pick-up in trade three to four years after limited or moderate sanctions have been lifted, but this evidence is tenuous because there are so few observations. Three out of five of the available coefficient estimates have positive signs but only one is statistically significant, LIM3*4 in 1995.
As noted, a high proportion of the variation in bilateral trade flows is explained by the factors in the model, so the lack of evidence for "lagged effects" probably does not stem from poorly defined variables. Nevertheless, in order to test for this possibility, and also because there are no observations in many instances, we tried redefining the lag variables in various ways.9 These results are broadly similar to those described above, suggesting that sanctions can have a lingering impact under some circumstances but that this is not a general effect. The failure to find lingering effects may be a product of the highly aggregated nature of the data used in our gravity model. Sanctions might have more pronounced long-term effects for certain types of products (e.g., sophisticated equipment that requires after service) or for particular sectors (e.g., infrastructure) than for the exporters as a group.
The Impact on OECD and US Exports
Keep in mind that all of the results described above apply to total bilateral trade, the sum of exports and imports between each pair of countries. Data for exports alone are also available in the data set, but only for 1995. In order to narrow the focus to the impact on the exports of the major users of sanctions, we also ran the regression using the logarithm of OECD exports and then of US exports alone as the dependent variable. The latter regression excludes all the pairs that do not have the United States as the exporter, reducing the number of observations to 84.
The results for OECD exporters are similar to those for the broader sample using total bilateral trade, except that the BLOC dummy is not significant and even has a negative sign. Of more interest, the coefficient on the limited sanctions variable is much greater than before and the impact of extensive sanctions is somewhat less (see Table 6). The limited sanction dummy variable is also statistically significant at the 99 percent level, which gives us much greater confidence in interpreting the results. The modest and extensive sanction variables are also still highly significant while the lagged sanction variables are mostly insignificant statistically or do not have observations that can be tested.10 The OECD results suggest that extensive sanctions lower exports by 78 percent, while limited to modest sanctions reduce exports by 21 percent to 33 percent.
Interestingly, comparing these results to those for total bilateral trade (exports plus imports) suggests that, while limited sanctions have a relatively larger impact on exports than imports, extensive sanctions have exactly the opposite relative impact.11 This result may be explained by the fact that even the most comprehensive sanctions regimes typically allow humanitarian shipments of food and medical supplies. Given the mercantilist tendencies of most governments, it seems likely that exemptions for "humanitarian" exports are eagerly sought and loosely regulated, while import sanctions may be more enthusiastically enforced.
The results for the US sample are also broadly similar to those for the larger sample, but because the model has substantially fewer data points it performs less well than before (Table 7). The key gravity variables--size and distance--have the expected signs and are statistically significant at the 95 percent confidence level, as is the dummy for a common language. But per capita income, adjacency, and the bloc dummy are no longer significant. The coefficient for limited sanctions suggests they have an impact on US exports similar to that for other OECD exporters, but we cannot have great confidence in this result since the coefficient is no longer statistically significant. The coefficients for the extensive and moderate sanction variables are much larger for the US sample than either for the sample as a whole or for the OECD sample, suggesting that the United States imposes relatively broader or tougher sanctions that have a greater adverse impact on its exports than the sanctions of most other OECD countries.
Because the United States is by far the largest user of unilateral economic sanctions, one might expect to find more robust evidence in US export patterns for lingering effects of sanctions after they have been lifted. Unfortunately, there is too little data to thoroughly analyze this question. There are no observations in three of the six instances (MOD1*2, MOD3*4 and EXT3*4). In the other three cases, the coefficients are not statistically significant and only one (LIM1*2) has a negative sign.
Another argument frequently heard in the debate is that US competitors move in and capture the business when the United States imposes unilateral sanctions. One way to explore this hypothesis is by examining the country pairs with "positive residuals" in their regression equations. Positive residuals indicate cases where actual trade is higher than the model would predict. If the "business capture" argument is correct, one would expect to find positive residuals for observations that pair major industrial countries such as France, Germany, and Japan, for example, with sanctions targets such as Iran, Libya, and China. This is, indeed, the case with respect to Cuba where the "positive residuals" indicate that Belgium, Canada, France, Germany, Ireland, Italy, Mexico, the Netherlands, and Spain trade more with Cuba than expected given size, income, and distance. In addition, Australia, Canada, and Germany export more to China than predicted by the model.
Calculating US Exports Lost
Although the estimated coefficients in the tables represent the average impact of sanctions, they can be used to calculate a reasonable estimate of the overall magnitude of the impact of sanctions on US exports.12 This calculation suggests that US exports were $15 billion to $19 billion lower than they would have been if not for the direct and indirect effects of sanctions in place in 1995 (see Table 8). The estimated reduction in annual US exports to countries targeted by sanctions would be expected to continue as long as sanctions of similar intensity are in place. In fact, the impact probably would grow over time since, in the absence of sanctions, exports to these countries would normally rise as they increase their income levels.
Jobs and Wage Effects of Sanctions
The United States is now enjoying full employment, and in a full employment economy, lower exports do not spell an overall drop in employment. However, it does mean that fewer workers are employed in the export sector of the economy, and more workers are employed elsewhere. According to the most recent US Department of Commerce study (1996), in the year 1992, $1 billion of goods exported supported 15,500 jobs, both directly in the exporting firms and indirectly in their suppliers.13 Taking into account productivity growth, the figure in 1995 was probably about 13,800 jobs. If the $15 billion to $19 billion estimated reduction in US exports in 1995 was not offset by exports to other markets, the loss of jobs in the export sector (if not in the economy as a whole) was between 200,000 and 260,000 positions.14
Jobs in the export sector pay better than average wages. This has been demonstrated by the careful econometric work of Richardson and Rindal (1996), and by the US Department of Commerce (1996).15 Thus, even in a full employment economy, the loss of exports means a loss of wages -- the export sector wage premium. Taking into account both direct and indirect employment, the export sector wage premium is about 12 to 15 percent.16 In 1995, when the average wage in manufacturing was about $34,020, the premium paid by the export sector was about $4,080 per worker (12 percent of $34,020). What these figures mean is that, as a consequence of US sanctions, workers probably lost somewhere between $800 million and $1 billion in export sector wage premiums in 1995.
The US practice of using economic sanctions extensively has become a fixture of US foreign policy at least since President Carter (1977-1980). In some periods during the past twenty years, when the US economy did not enjoy full employment, and when jobs were not readily available, the loss of exports may have added to the unemployment rolls. But even if the loss of exports had zero effect on total employment, it certainly reduced the number of good paying jobs. If the next twenty years see the same frequent application of sanctions, the cumulative loss of wage premiums could exceed $20 billion (20 years times roughly $1 billion a year, not taking into account the rising annual loss of exports). This is a heavy cost.
Notes
1. See Gary Clyde Hufbauer, Jeffrey J. Schott, and Kimberly Ann Elliott, Economic Sanctions Reconsidered, second edition, revised (Washington: Institute for International Economics, 1990).
2. Two notable exceptions are, J. David Richardson, Sizing Up U.S. Export Disincentives (Washington: Institute for International Economics, , 1993); and National Academy of Sciences, Balancing the National Interest (Washington: National Academy Press, 1987).
3. The model used here is described in detail in Jeffrey A. Frankel, Regional Trading Blocs in the World Economic System (Washington: Institute for International Economics, forthcoming 1997). The model coefficients were calculated by Tess Cyrus, University of California Berkeley, using sanctions variables supplied by the Institute for International Economics.
4. Recall that a logarithm is the exponent applied either to the base 10 or to the base e to generate the data point in question (for example, GNP). In this study, logarithms are taken to the base e.
5. See Frankel (1997) for a thorough analysis of the various methods for measuring distance and how they affect the results. The method used here measures the "great-circle" distance.
6. The test involved running two sets of non-linear regressions, one which included the zero observations and one without. These produces almost identical results. This finding underpins our assertion that eliminating the zero trade observations for the ordinary least squares regression does not substantially distort the results.
7. The percentage change in trade is calculated by taking the exponent of the coefficient value for the dummy and subtracting 1. For example, from Table 2, the coefficient for limited sanctions (LIM) is -0.331. The value of the natural number e taken to the exponent -0.326 is 0.722. This indicates that bilateral trade was only 0.722 times as large, or 27.8 percent lower, between two countries owing to the limited sanctions than it would have been if the sanctions were not in place.
8. percent but the coefficient is not significant at standard levels and the coefficient on the same variable is not statistically significant in either 1990 or 1985.
9. We tried three different variants on the lagged effect variables shown in Tables 2, 3, and 4. First, we reduced the six lag variables to just three, one for each sanctions level regardless of when in the previous four years the sanctions were lifted. Second, we retained the one-to-two and three-to-four-year lag structure but eliminated the levels of intensity. In other words, the six lag variables are collapsed to two: one for sanctions of any intensity lifted in the past one to two years, and one for any sanctions lifted three to four years previously. A single lag variable, representing any sanction of any intensity lifted in the previous four years, did show a negative estimated impact on trade in all three years (1985, 1990, 1995) but was never statistically significant.
10. We also ran the regression using the alternative lag variable definitions, but the coefficients were still mostly insignificant or had positive signs suggesting a pick-up in exports some years after sanctions have been lifted.
11. Additional analysis of OECD imports alone confirmed this result.
12. Because the coefficients represent the average impact of sanctions, however, great care should be taken in interpreting the country-by-country results.
13. US Department of Commerce, US Jobs Supported by Exports of Goods and Services, (Washington D.C.: US Department of Commerce, November 1996).
14. This estimate will be exaggerated to the extent that a portion of the goods previously shipped to sanctioned countries was diverted to other markets.
15. J.D. Richardson and K. Rindal, Why Exports Matter: More! (Washington, D.C.: Institute for International Economics and The Manufacturing Institute, 1996).
16. US Department of Commerce (1996). Richardson and Rindal (1996) found that the export wage premium was about 16 percent for workers employed directly in producing manufactured exports.
Appendix 1: Countries in Data Set
Algeria | Hungary | Paraguay |
Angola | Iceland | Peru |
Argentina | India | Philippines |
Armenia | Indonesia | Poland |
Australia | Iran | Portugal |
Austria | Iraq | Romania |
Belgium | Ireland | Russia |
Bolivia | Israel | Saudi Arabia |
Brazil | Italy | Serbia/Montenegro |
Bulgaria | Japan | Singapore |
Canada | Kenya | Somalia |
Chile | South Korea | South Africa |
China | Kuwait | Spain |
Colombia | Libya | Sudan |
Cuba | Lithuania | Surinam |
Czechoslovakia | Macedonia | Sweden |
Denmark | Malawi | Switzerland |
Ecuador | Malaysia | Syria |
Egypt | Mexico | Taiwan |
El Salvador | Morocco | Thailand |
Ethiopia | Myanmar (Burma) | Tunisia |
Finland | Nepal | Turkey |
France | Netherlands | United Kindgom |
Gambia | New Zealand | United States |
Germany | Nicaragua | Uruguay |
Ghana | Nigeria | Venezuela |
Greece | North Korea | Vietnam |
Guatemala | Norway | Zimbabwe |
Haiti | Pakistan | |
Hong Kong | Panama |
Table 1: Glossary for sanctions variables
Dummy Variable | Explanation |
|
|
LIM | Limited financial, cultural, travel, or trade restrictions |
MOD | Broader trade or financial restrictions |
EXT | Comprehensive trade and financial restrictions |
LIM1*2 | Limited sanctions lifted one to two years ago |
MOD1*2 | Moderate sanctions lifted one to two years ago |
EXT1*2 | Extensive sanctions lifted one to two years ago |
LIM3*4 | Limited sanctions lifted three to four years ago |
MOD3*4 | Moderate sanctions lifted three to four years ago |
EXT3*4 | Extensive sanctions lifted three to four years ago |
Table 2: 1985 Results, All Countries
|
||||
Variable | Estimated coefficient |
Standard error |
t-statistic | |
|
||||
C (constant) | -9.334 | 0.438 | -21.306 | -21.306 |
Standard gravity model variables | ||||
LGNP | 0.770 | 0.015 | 52.117 | * |
LGNPPC | 0.209 | 0.018 | 11.907 | * |
LDIST | -0.779 | 0.035 | -22.525 | * |
ADJ | 0.507 | 0.156 | 3.244 | * |
LANG | 0.488 | 0.075 | 6.492 | * |
BLOC | 0.491 | 0.170 | 2.889 | * |
Sanctions dummy variables | ||||
LIM | -0.326 | 0.121 | -2.692 | * |
MOD | -0.787 | 0.295 | -2.672 | * |
EXT | -2.424 | 0.432 | -5.611 | * |
LIM1*2 | 0.209 | 0.528 | 0.397 | |
MOD1*2 | -0.514 | 0.646 | -0.796 | |
EXT1*2 | na | na | na | |
LIM3*4 | -0.135 | 0.335 | -0.403 | |
MOD3*4 | 0.691 | 0.644 | 1.073 | |
EXT3*4 | -1.597 | 0.744 | -2.148 | ** |
* Significant at the 99 percent confidence level or better.
**Significant at the 95 percent confidence level or better.
Number of observations | 2,472 |
Standard error of regression | 1.286 |
Adjusted R2 | 0.691 |
na means there are no observations in the sample with this type of sanction.
Table 3: 1990 Results, All Countries
|
||||
Variable |
Estimated
coefficient |
Standard
error |
t-statistic | |
|
||||
C (constant) | -8.871 | 0.406 | -21.856 | |
Standard gravity model variables | ||||
LGNP | 0.794 | 0.014 | 56.352 | * |
LGNPPC | 0.086 | 0.015 | 5.611 | * |
LDIST | -0.704 | 0.034 | -20.547 | * |
ADJ | 0.718 | 0.149 | 4.825 | * |
LANG | 0.620 | 0.073 | 8.503 | * |
BLOC | 0.523 | 0.148 | 3.541 | * |
Sanctions dummy variables | ||||
LIM | -0.520 | 0.124 | -4.205 | * |
MOD | -0.238 | 0.135 | -1.762 | ** |
EXT | -2.288 | 0.371 | -6.163 | * |
LIM1*2 | -0.986 | 1.280 | -0.771 | |
MOD1*2 | -0.783 | 1.280 | -0.612 | |
EXT1*2 | -0.878 | 0.905 | -0.970 | |
LIM3*4 | 0.600 | 0.573 | 1.047 | |
MOD3*4 | na | na | na | |
EXT3*4 | na | na | na |
* Significant at the 99 percent confidence level or better.
**Significant at the 95 percent confidence level or better using a one-tailed test.
Number of observations | 2,628 |
Standard error of regression | 1.279 |
Adjusted R2 | 0.721 |
na means there are no observations in the sample with this type of sanction.
Table 4: 1995 Results, all countries
|
||||
Variable | Estimated coefficient |
Standard error |
t-statistic | |
|
||||
C (constant) | -8.050 | 0.360 | -22.352 | |
Standard gravity model variables | ||||
LGNP | 0.811 | 0.013 | 62.286 | * |
LGNPPC | 0.091 | 0.014 | 6.317 | * |
LDIST | -0.856 | 0.032 | -26.401 | * |
ADJ | 0.570 | 0.142 | 4.018 | * |
LANG | 0.787 | 0.069 | 11.479 | * |
BLOC | 0.175 | 0.128 | 1.362 | |
Sanctions dummy variables | ||||
LIM | -0.148 | 0.119 | -1.246 | |
MOD | -0.374 | 0.139 | -2.691 | * |
EXT | -2.510 | 0.227 | -11.053 | * |
LIM1*2 | -0.156 | 0.148 | -1.058 | |
MOD1*2 | -0.990 | 0.613 | -1.615 | *** |
EXT1*2 | -0.233 | 0.244 | -0.955 | |
LIM3*4 | 0.653 | 0.283 | 2.309 | ** |
MOD3*4 | na | na | na | |
EXT3*4 | na | na | na |
* Significant at the 99 percent confidence level or better.
**Significant at the 95 percent confidence level or better.
***Significant at the 90 percent confidence level or better.
Number of observations | 2,714 |
Standard error of regression | 1.225 |
Adjusted R2 | 0.761 |
na means there are no observations in the sample with this type of sanction.
Table 5: Impact of Sanctions on Bilateral Merchandise Trade
change in trade |
||||
|
||||
Level of sanctions imposed in base year | 1985 | 1990 | 1995 | Average |
|
||||
All countries, exports plus imports | ||||
Limited (LIM) | -27.8 | -40.6 | -13.8a | -27.4 |
Moderate (MOD) | -54.5 | -21.2 | -31.2 | -35.6 |
Extensive (EXT) | -91.1 | -89.9 | -91.9 | -91.0 |
OECD countries, exports only | ||||
Limited (LIM) | na | na | -21.5 | na |
Moderate (MOD) | na | na | -33.1 | na |
Extensive (EXT) | na | na | -78.0 | na |
a. The coefficient on this variable was not statistically signficant at the usual levels.
na means no regression was run for OECD exports in this year.
Table 6. 1995 Results, OECD countries, exports only
Variable | Estimated coefficient |
Standard error |
t-statistic | |
|
|
|
|
|
C (constant) | -9.816 | 0.381 | -25.774 | |
Standard gravity model variables | ||||
LGNP | 0.817 | 0.013 | 63.237 | * |
LGNPPC | 0.131 | 0.017 | 7.968 | * |
LDIST | -0.850 | 0.031 | -27.387 | * |
ADJ | 0.342 | 0.146 | 2.348 | ** |
LANG | 0.836 | 0.079 | 10.559 | * |
BLOC | -0.012 | 0.097 | -0.119 | |
Sanctions dummy variables | ||||
LIM | -0.242 | 0.093 | -2.595 | * |
MOD | -0.402 | 0.155 | -2.592 | * |
EXT | -1.513 | 0.230 | -6.586 | * |
LIM1*2 | -0.048 | 0.178 | -0.217 | |
MOD1*2 | na | na | na | |
EXT1*2 | 0.470 | 0.308 | 1.525 | |
LIM3*4 | 0.703 | 0.252 | 2.792 | * |
MOD3*4 | na | na | na | |
EXT3*4 | na | na | na |
* Significant at the 99 percent confidence level or better.
**Significant at the 95 percent confidence level or better.
Number of observations | 2,118 |
Standard error of regression | 1.055 |
Adjusted R2 | 0.812 |
na means there are no observations in the sample with this type of sanction.
Table 7: 1995 Results, US Exports
Variable | Estimated coefficient |
Standard error |
t-statistic | |
|
||||
C (constant) | -12.448 | 3.219 | -3.867 | |
Standard gravity model variables | ||||
LGNP | 0.856 | 0.088 | 9.772 | * |
LGNPPC | 0.137 | 0.117 | 1.172 | |
LDIST | -0.643 | 0.319 | -2.013 | ** |
ADJ | 0.550 | 1.394 | 0.395 | |
LANG | 0.588 | 0.287 | 2.046 | ** |
BLOC | 0.412 | 1.050 | 0.393 | |
Sanctions dummy variables | ||||
LIM | -0.234 | 0.332 | -0.705 | |
MOD | -1.138 | 0.514 | -2.214 | ** |
EXT | -3.452 | 0.631 | -5.470 | * |
LIM1*2 | -0.208 | 1.046 | -0.199 | |
MOD1*2 | na | na | na | |
EXT1*2 | 1.880 | 1.098 | 1.712 | *** |
LIM3*4 | 0.322 | 0.591 | 0.545 | |
MOD3*4 | na | na | na | |
EXT3*4 | na | na | na |
* Significant at the 99 percent confidence level or better.
**Significant at the 95 percent confidence level or better.
***Significant at the 90 percent confidence level or better.
Number of observations | 84 |
Standard error of regression | 1.009 |
Adjusted R2 | 0.809 |
na means there are no observations in the sample with this type of sanction.
Table 8: US exports lost due to economic sanctions in place in 1995
Target country |
Sanctions level 1995 |
Actual exports 1995 |
Estimated reduction in US exports using coefficients for OECD exportse |
Estimated reduction in US exports using coefficients for US exportsf |
Estimated reduction in US exports using coefficients for US exportsf |
|
|||||
Angola | MOD | 260 | 128.3 | 551 | |
Bulgaria | LIM | 132 | 36 | 36 | * |
Burma | MOD | 16 | 8 | 34 | |
China | LIM | 11,749 | 3,213 | 3,213 | * |
Cubaa | EXT | 6 | 1,094 | 1,094 | |
Czech | LIM | 556 | 152 | 152 | * |
Ecuador | LIM | 1,538 | 421 | 421 | * |
Guatemala | LIM | 1,652 | 452 | 452 | * |
Hungary | LIM | 296 | 81 | 81 | * |
India | LIM | 3,296 | 901 | 901 | * |
Indonesia | LIM | 3,395 | 929 | 929 | * |
Iran | EXT | 238 | 842 | 2,526 | |
Iraqb | EXT | - | 1,992 | 1,992 | |
Libyab | EXT | - | 1,513 | 1,513 | |
Nigeria | LIM | 602 | 165 | 165 | * |
North Koreac | EXT | 5 | 185 | 185 | |
Pakistan | MOD | 935 | 461 | 1,983 | |
Peru | LIM | 1,775 | 485 | 485 | * |
Poland | LIM | 776 | 212 | 212 | * |
Romania | LIM | 256 | 70 | 70 | * |
Russia | LIM | 2,651 | 725 | 725 | * |
Serbia & Montenegrod |
EXT | - | 675 | 675 | |
Sudan | MOD | 44 | 22 | 93 | |
Syria | MOD | 223 | 110 | 473 | |
The Gambia | LIM | 7 | 2 | 2 | * |
Vietnam | LIM | 253 | 69 | 69 | * |
TOTAL | 14,942 | 19,031 |
a. Estimated 1995 exports to Cuba calculated as (US share of 1995 OECD exports to the Dominican Republic)
multiplied by (1995 OECD exports to Cuba)
b. Estimated 1995 exports to Iraq, Libya calculated as (US exports in last year before sanctions) multiplied by
(world exports 1995/world exports last year before sanctions).
c. Estimated 1995 exports to N. Korea calculated as (US share of 1995 OECD exports to S. Korea) multiplied by
(1995 OECD exports to N. Korea)
d. Estimated 1995 exports to Serbia calculated as estimated 1995 US exports to all of Yugoslavia less observed
exports to Croatia and Bosnia Herzegovina.
e. Calculated as [(actual 1995 exports)/exponent of coefficient from Table 6)] - actual 1995 exports.
f. Calculated as [(actual 1995 exports)/exponent of coefficient from Table 7)] - actual 1995 exports.
*The limited (LIM) coefficient for US exports-only in 1995 (Table 7) is not statistically significant, so the OECD
exports-only coefficient is used in its place.
Related Documents
Document
Appendix and Tables (102.84 KB)