UNIVERSITY OF GRONINGEN
Does Trade Transmit Effects of Armed Conflicts to
Asset Markets?
Analyzing Trade Transmission Channels of Effects of the 2003 Iraq
Invasion
Kasper Riezebos1
Supervisors: Prof. Dr. W. Bessler & Dr. M.J. Gerritse
Abstract
This paper analyzes if effects of armed conflict on asset markets are moderated by pre-war trade relations. In particular, the paper tries to understand if pre-war exports and imports transmit effects between war on the one side, and stock market indices, defense stocks, currency exchange rates, government bonds, and commodities on the other side. The research indeed finds that pre-war trade transmits effects of war on asset markets, in particular on stock markets.
Key Words: Armed conflict, Transmission channels, Asset markets, Trade JEL Classifications: F42, F51, G11, G18
1 Student MSc. International Economics & Business and MSc. Finance, Faculty of Economics & Business, University of
1
I.
Introduction
It is important to policy makers and asset market managers to understand effects of events with large economic consequences. It might be even more important to understand via what transmission channels an event has those consequences. In this thesis the effects of a large event; armed conflict, are assessed. More specifically, this thesis analyzes if trade transmission channels exist for the effect of armed conflict on asset markets. Due to data constraints I focus on one particular case; the 2003 US-led coalition invasion of Iraq, which at that time was controlled by the Saddam Hussain regime.
The Iraq 2003 conflict had its roots partially in the Iraq’s invasion of Kuwait in 1990. One of the consequences of this conflict were that trade and financial sanctions were imposed on Iraq. Since this resulted in food shortages in Iraq, from 1996 onwards the Iraqi government was allowed to export up to $5.256 billion worth of oil every 6 months (Katzman, 2003) in exchange for food. Consequences of these trade restrictions of the UN food-for-oil program were that it benefited certain countries more than others2 and that the potential of Iraq’s oil production could not be fully exploited: only 17 out of 80 discovered oil fields had been developed under Saddam Hussein’s administration (Jones, 2003). Following the 9/11 attacks and the so-called ‘war on terror’, tensions rose again between Iraq and the US. In the end the conflict could not be resolved politically and resulted in the invasion of Iraq in 2003. This invasion was also known as ‘Operation Iraqi Freedom’ and was a “US-led coalition military operation in Iraq” (Dale, 2011) consisting of American, British, Australian, and Polish troops, aimed at “disarming Iraq of weapons of mass destruction”3. The operation lasted 21 days; it started on the 19th of March and ended on the 13th of April 2003. The invasion began with 2 days of aerial bombardments, followed by a ground assault on the 21st of March. During the course of the attack, especially in the first week, most oil wells were captured without being sabotaged by the Iraqi regime (Fontenot, Degen and Tohn, 2004)4. On the 9th of April Bagdad was completely in coalition hands. During and before the invasion, it was possible to buy futures paying out if Saddam Hussain was ousted before a certain date. This makes this invasion of particular interest for academic research, as it is possible to measure expectations regarding the outcome of the invasion on a day to day basis.
2 For an example of the reasoning who would profit from keeping Saddam Hussain in power, see an article of Scatterlee,
of the Heritage Foundation.
3 According to the official White House news release.
2
The Iraq invasion of 2003 had consequences on asset markets. Stock markets rallied in the US and a number of other countries, while other countries saw less pronounced effects on stock markets (Wolfers and Zitzewitz, 2009). War causing an increase in stock markets might seem counterintuitive since the violence associated with armed conflict usually destroys value. This raises the question: what determines the effects of armed conflict on asset markets? Are some countries hurt more than others, and if so, what transmission channels decide how asset markets in various countries are affected?
This thesis will continue as following: the next section will discuss existing literature regarding this topic. Section 3 will explain the research method and the data used. Section 4 will present the output of the analysis and will subsequently discuss the results and the corresponding robustness tests. The thesis is concluded in section 5.
II. Literature review
To be able to answer the question in the previous section, it is important to define the key definitions and discuss existing research on this topic. The Department of Peace and Conflict Research of the University of Uppsala defines an armed conflict as “a contested incompatibility that
concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths in one calendar year”. 5 The term war will be used interchangeably with armed conflict in this this thesis. Armed
conflicts are different from terrorism as terrorism is defined by the Patriot Act in the United States as “and act dangerous to human life (…) to be intended to: intimidate a coerce a civilian population;
influence the policy of a government by intimidation or coercion (…) mass destruction, assassination or kidnapping” (Hesterman, 2014). Terrorism thus mainly differs from armed conflict in the sense
that to define an event as an armed conflict it requires more casualties to have fallen. Thereby, the intention to intimidate is more important by defining terrorism. The remainder of this thesis will focus on the case of the invasion of Iraq in 2003, which was by all means an armed conflict satisfying the above definition of the Department of Peace and Conflict Research of the University of Uppsala. Armed force was used in the conflict and main belligerents were the governments of the United States, United Kingdom, Australia, Poland, and Iraq. Estimated battle-related deaths are between 10,800 and 15,100 on the Iraqi side alone (Conetta, 2003).
3
Previous research finds that political conflicts have an effect on a range of economic parameters, such as economic growth, inflation, and currency exchange rates. For example, conflicts affect financial development through a drop in value of the local currency and decrease financial sector governance in a country experiencing armed conflict (Addison et al., 2002). Tax income, inflation, investments, economic growth (Gupta et al., 2003) and money (McCandless, 1996) are also affected. Abadie and Gardeazabal (2003) corroborate the negative effect of an armed conflict on economic growth in a case study of the conflict in the Basque Country. Bilateral trade seems to be influenced as well by “broad political relations of amity and enmity between nations” (Pollins, 1989). Similar findings regarding trade and conflicts are presented by Karstner (2007), who measures what the effects are of UN voting patterns on trade. Karstner finds that political interest often correlates with bilateral trade.
A reversed causality between economic parameters and conflict is also possible. There could be reversed causality between trade and conflict; more trade would make political conflicts less attractive, since costs of conflict would have increased (Polachek, 1980; Polachek, 1992). However, this claim of “trade brings peace” is disputed, other results indicate that trade does not bring peace but peace and trade are correlated because of a “simultaneity bias” (Keshk et al., 2004). Other research shows that financial crises predict political crises (Bertin, Ohana and Strauss-Kahn, 2014) and that higher oil prices are linked to more aggressive behavior by oil producing states (Hendrix, 2014). Armed conflict thus affects economic factors, and is affected by economic factors itself as well.
4
Armed conflicts also affect stock markets, as this paragraph will show. Stock markets are defined as “equity markets” or “markets for trading equities” by the American Stock Market Nasdaq. Abadie and Gardeazabal (2003) prove that stock markets in the case of the Basque Country conflict reacted positively to the start of ceasefire talks and reacted negatively when they failed. This indicates that the negative economic consequences of armed conflict are also reflected in the perceived value of companies in countries that are involved in the conflict. Thereby, a considerable body of research is conducted to examine the effects on stock markets of a more specific case of armed conflicts; terrorist attacks. Stock markets are found to be negatively affected if a terrorist attack occurs (Karolyi and Martell, 2006; Spagnolo et al., 2008; Eldor et al., 2004). Acts of terrorism thus seem to have a significant effect on stock markets. Whereas terrorism, as defined before, is usually not country-wide violence over a longer period, the Iraq invasion of 2003 took several weeks and affected a larger area than most acts of terrorism. Therefore, one would also expect armed conflicts as the Iraq invasion of 2003 to be reflected in asset market movements.
Western asset markets are indeed affected by armed conflict (Schneider and Troeger 2006). Two researches focusing on the Israeli-Palestinian conflict find that the outbreak of armed conflict negatively affected asset markets, while peace talks and a successful offensive led to a stock market increase (Zussman et al., 2006; Kollias et al., 2010). The causal relationship between conflict and asset prices is corroborated in a study of 101 internal and inter-state conflicts by Guidolin and Ferrara (2005); a significant number of conflicts has an effect on “stock market indices, exchange rates, oil and commodity prices”. Surprisingly, and contrary to the researches mentioned before, the outbreak of a conflict led to a rise in stock market indices of the United States, United Kingdom, and France. Other research partially confirms the finding that the outbreak of conflict leads to increasing stock market indices, whereas before a war, increased likelihood of a conflict decreased stock prices (Brune, Hens, Rieger and Wang, 2015). During World War II, British stock markets displayed significant positive reactions to good news and significant negative reactions to bad news (Urquhart, 2015). A short overview of the most important effects discussed until now is shown in figure 1, in which the question mark for the relation between conflict and stock markets denotes the difference in reaction of stock markets depending on threat of conflict, outbreak of conflict and good or bad news regarding the conflict. The question mark for the reaction of defense stocks depict the uncertainty regarding the effects of armed conflict on defense stocks.
5
Prediction markets, also called “information markets” or “event futures” are “markets where
participants trade in contracts whose pay-offs depends on unknown future events” (Wolfers and
Zitzewitz, 2004). The prediction market used in the Wolfers and Zitzewitz (2009) paper is the Saddam Security, which will be explained into more detail in section 3.1. The research of Wolfers and Zitzewitz also analyzes which countries and industries are most sensitive to this conflict. They find that sensitivity to conflict differs across countries and industries. Following this paper, Amihud and Wohl (2004) use the same Saddam Security to test the effect of war expectations of financial markets on bond prices, exchange rates, and stock market indices. They find that an increasing probability that Saddam would be ousted strengthened the Dollar against the Euro, caused an increase in stock prices and resulted in lower oil prices. They also found that oil prices increased before the conflict, possibly because of supply uncertainty in the case of a war. This supply uncertainty effect is also acknowledged by Elwanger (2015) who finds that supply uncertainty is a potential reason for oil prices to be very volatile during a conflict involving one or more oil supplying nations. If investors are not sure that future oil demand can be met, uncertainty increases and oil prices will be more volatile. While oil production could be hurt by armed conflicts, defense contractors could be the ones to benefit from war. This would be an industry effect instead of market wide effects. Although it might seem straightforward that the defense industry will profit because demand for defense products will increase, this case is still indistinct. Kaen (2012) does not find a strong relationship between war and defense stocks, while Berrebi and Klor (2010) find that acts of terrorism have a strong positive effect on the stocks of defense related companies.
6
power in Bagdad on a date specified before. Using the daily prices of this security, one can measure what the markets thought the chance was of an invasion in Iraq, and once the invasion started, how successful markets thought the invasion would be. Thus, for the cases with relevant data available, armed conflict is measurable.
As showed in the literature discussed, existing research supports the claim that there is an effect of armed conflict on currency exchange rates, commodity prices, bond prices, and stock markets of a number of countries directly and not directly involved in the conflict. But how is this effect transmitted to asset markets in different countries? In a paper researching the volatility transmission in international stock markets, Koutmos and Booth (1995) find that there is a ‘transmission mechanism’ for volatility and price spillovers between different stock markets. Different stock markets did react differently to the same news. It could thus very well be that stock markets of different countries display diverse reactions to the same war news. And indeed, Wolfers and Zitzewitz (2009) find that national stock markets react diverse to expectations about the Iraq invasion. Existing research as the paper of Wolfers and Zitzewitz, however, does not make clear what specific transmission channels could cause these different asset market reactions to armed conflict. As showed in the literature review, trade influences conflict and vice versa, but transmission channels which cause these different asset markets reactions to war are still unclear.
7
Do pre-‘armed conflict’ exports and imports transmit the effect of armed conflict to asset markets of different countries?
As highlighted in the previous section, through the trade transmission channels, asset markets in different nations could be differently affected by the Iraq invasion. In the case of the 2003 Iraq invasion it would make sense from an international economics perspective that countries trading more with Iraq would see relatively negative reactions on stock markets. I expect this to be the case since businesses from those countries would lose demand of their goods and supply of raw materials. However, since some countries were dependent on Iraqi oil, countries importing more from Iraq could see the war as a chance of lifting the sanctions, and oil imports in the future being more accessible. To test these expectations, the hypothesis is formulated as follows:
H0: The effects of expectations regarding the outbreak or outcome of armed conflict are not
moderated by exports and imports between the country experiencing war and the country in which asset markets are based.
III. Data and research method
To answer the research question proposed in the previous section, this section will first discuss the data used to test the hypothesis.
3.1 Conflict measure
The conflict parameter that is used in the regressions is the so-called ‘Saddam Security’ which I have obtained from TradeSports.com via prof. Zitzewitz and Mr. Bernstein6. The Saddam Security has data from September 2002 until the end of the Iraq invasion in 2003, consisting of several different Saddam Securities: for December 2002, March 2003, April 2003, May 2003, June 2003 and July 2003. These securities payed $10 if Saddam was ousted of the center of Bagdad before the last day of these months.
According to Wolfers and Zitzewitz (2009); monthly trading volumes of the Saddam Securities were just over 10,000 dollars in the period before the war took place, and increased during the war. Trade over the total period was approximately $1.2 million. In figure 2 the price development of the whole period the June Saddam Security traded is shown. Figure 3 depicts the
6 Many thanks to Mr. Bernstein, former CEO of ‘tradesports.com’ and prof. Zitzewitz for providing me the relevant
8
Figure 2: June Saddam Security price development
Note: The price of the Saddam Security is multiplied by ten to show the price on a zero to 100 scale instead of a zero to
ten scale. Total observations number 1,000 over approximately seven months and there are multiple observations per day. Observations are from September 25th until April 11th.
Figure 3: volatility of the Saddam Security
Note: Price changes are relative to the price of the last observation. Total observations number 1,000 over approximately
seven months and are multiple times per day. Observations are from September 25th until April 11th.
40 60 80 100
Price Saddam Security In US$ * 10.
01oct2002 01dec2002 01feb2003 01apr2003
Date -.2 0 .2 .4 Relative changes Saddam Security
01oct2002 01dec2002 01feb2003 01apr2003
9
The analysis in this paper makes use of the June Saddam Security, because this variable has the most frequent observations, and using the June Security ensures both enough observations and the data continuing until the end of the war (13th of April 2003)7. As seen in figure 2, prices start at approximately 93 dollars on March 19, and rise during the start of the war. Then a period occurs in which the invasion was not as successful as apparently expected by the market, until the 1st of April. From then on the expectations of a successful end of the war rose again, until the 10th of April, the day after the fall of Bagdad to US coalition troops.
The analysis uses the relative first differences of the June Saddam Security, to be able to examine the effect of the variable on a number of dependent variables, such as stock market indices. The relative first differences formula is computed as:
𝑆
̃ =
𝑖 𝑆𝑖−𝑆𝑖−1𝑆𝑖−1 (1)
In which Si is the value of the Saddam Security on day i, and Si-1 is the value of the Saddam Security on day i-1. Since the Saddam Security gives prices several times a day, and all other variables only give prices once a day, I take the daily average price of the June Saddam Security. Data from September 25, 2002 until the 11th of April 2003 is used because that is the period the Saddam Security sold. Trade of the June Saddam Security stopped after the 11th of April, a day after the fall of Bagdad. All observations with missing data are deleted; during the war there are no missing observations for the Saddam Security. There are only missing observations for the Saddam Security before February 1st. Deleted are also observations of the Saddam Security with more than 2 days’ difference between the last observation.
A potential problem could be that there is an endogeneity bias between the financial markets and the Saddam Security. Wolfers and Zitzewitz (2009) address these concerns by using the Saddameter of Saletan to control for a potential effect of financial markets on the prices of the Saddam Security. This estimation using the Saddameter concludes that any potential effect of financial markets on the Saddam Security “is less than 10%”.
During the whole sample there were almost 1000 observations, almost 200 a day. In December 2002 there were relatively less trades as clearly visible in figures 2 and 3. Volatility using all observations available is depicted in graph 3. Volatility is relatively high if the price of the
7 According to Encyclopedia Britannica, the complete Iraq War took until 2011. The initial invasion phase of the war was
10
Saddam Security goes down, while volatility is lower if the price of the Saddam Security is increasing or stable.
3.2 Dependent variables
For the year 2003, several stock market indices are derived from DataStream. The closing prices of the indices of the United States, United Kingdom, Russian Federation, China, France, Germany and Japan are analyzed. In line with Wolfers and Zitzewitz, the natural logarithm of the stock index prices is taken and then the first differences from day to day are analyzed. For the indices with corresponding index numbers see appendix A. All stock market indices have 19 observations during the war, except Japan, which has 18 observations. Total observations used for the stock market indices range from 83 to 90 observations per stock index. Potentially problematic and not accounted for in this research is that the asset markets are placed in different markets and different time-zones. This means that they are not active at the same time, and might react on the same war news on different days.
The relation between the Iraq invasion of 2003 and the commodities oil and gold is also researched. Since there are several kinds of oil indices, only the most relevant oil index is extracted from DataStream for the year 2003; the Dubai Crude Oil index, which is an oil index for Persian Gulf oil. To control for eventual effects of Venezuelan strikes on oil prices, I add a Venezuela bond index of Citigroup to the analysis. This is measured in dollars, and extracted from Thomson Reuters DataStream.
The HUI Index from the New York Stock Exchange for the year 2003 is derived from DataStream to analyze the effect of the Iraq invasion of 2003 on gold prices. Again, for both gold and oil, the natural logarithm of the commodity index prices is taken, and then the first differences from day to day are analyzed. For both commodities there are 89 observations. The analysis uses the Dubai Crude oil spot price. This is done to measure the effect on the local oil prices. The effect of war on future oil prices could be argued to be higher, as future oil production could be damaged by armed conflict. However, the difference between spot and future prices is not expected to be significant, as a paper by Bhardwaj, Gorton and Rouwenhorst (2015) shows.
11
dependent variable, the natural logarithm of the currency prices is taken, and then the first differences from day to day are analyzed. For the currencies and the corresponding index numbers, see appendix B.
The progress of the Iraq invasion could also have its effects on government bonds. To measure this, 5 bond indices of the following countries are used: the United States, United Kingdom, Japan, France, and Germany. For China and Russia there were unfortunately no bond indices available for the relevant time period. For the countries with their corresponding index numbers; see appendix C. The data used are 10-year Government Benchmarks extracted from Thomson Reuters DataStream. The natural logarithm of the government bond indices is taken, and then the first differences are analyzed.
Defense company stocks are not an asset class, but part of an industry. Defense company stocks traded on stock markets are already measured to a certain extend in the analysis of stock market indices. However, the defense industry could benefit from armed conflict, and the measured effects of armed conflicts on defense industry stocks are ambiguous until now. Therefore, to be able to measure the potential effect of armed conflict on the stock prices of defense companies, a number of different firms participating in the production of military goods and services are included in a regression. For all countries except Russia there was data available for publicly traded defense companies in 2003. The data used are again closing prices. Considering the defense industry analysis as well, the natural logarithm of the defense companies’ stock prices is calculated, and then the first differences from day to day. The data for these defense companies is obtained from DataStream. For the defense companies in the research with corresponding index numbers and countries, see appendix D.
3.3 Transmission channels
There are several possibilities for the effects of war to be channeled to asset markets. In this section the transmission channel data of main interest is discussed, as well as the rationale behind some control transmission channels and the data sources of these transmission channel variables. The transmission channels of main interest are ‘exports’ and ‘imports’ of Iraq. Potential control transmission channels discussed are Foreign Direct Investment (FDI), geographical and cultural distance, political relations between countries, and participation in the Iraq invasion of 2003.
12
1998 until 2002 are taken. The trade data is thus lagged data, as the invasion took place in 2003. ‘Exports’ is trade from Iraq to other nations whereas ‘imports’ is trade from other nations to Iraq. Exports and imports could be skewed because of Gross Domestic Product (GDP) differences between countries, therefore; exports and imports are divided by the GDP of the country in question. The data on GDP is extracted from the World Bank Database. There is one shortcoming in this analysis; due to the trade embargo Iraq could choose which country was awarded with its oil exports. Therefore, trade and political relations could be correlated and it is crucial to check for political relations and a number of other transmission channels to verify if the exports and imports transmission channels really exist.
A first control transmission channel is participation in the invasion. If an invasion seems to be successful, stock market participants from the invading nation could see the possibility of a long and expensive war decrease and future business opportunities arise. This would cause stock markets of invading nations to be more optimistic if an invasion proceeds successfully. Wolfers and Zitzewitz (2009) analyze the effect of participation in the invasion, and find that this has no significant effect on the stock markets of those countries. Two countries of my sample participated in the Iraq invasion: the United Kingdom and the United States of America. Germany, France and Russia officially voiced opposition to an invasion of Iraq8. Japan did not participate in the invasion, but was a member of the “coalition of the willing” which implies that Japan approved the invasion9. China disproved nor supported the invasion.
A second control transmission channel is geographical distance, since this could have an effect on exports and imports. Thereby, stock market participants could be fearful of the proximity of the war. Studying the spillover effects of sovereign debt ratings, Ferreira and Gama (2007) find that geographical closeness strengthens the effects of a downgrade on another country. Therefore, I use geographical distance as a control variable in my analysis. The control variable for distance is derived from the Centre for Prospective Studies and International Information (CEPII) database (Mayer and Zignago, 2011) and measures geographical distance in kilometers. All seven measured countries are different from Iraq in terms of language and ethnicity, thereby Iraq was never a colony of any of the seven nations included in the research, according to this same CEPII database. Cultural distance is thus treated as being equal between Iraq and the other nations. For the currency exchange
8 France, Germany and Russia said prior to the war that “they would not allow the UN to pass a resolution authorizing
war against Iraq”, according to BBC News. Extracted from http://news.bbc.co.uk/2/hi/middle_east/2821145.stm .
9 Japan is in de official list the White House posted, consisting of forty-nine countries publicly committed to the
13
rates, the geographical distance for Germany and France is averaged, because the Euro is the currency of both nations.
A third control transmission channel could be Foreign Direct Investment (FDI). If firms of a certain nation have invested relatively more in a country, stock market reactions of this country to the outbreak of a war would be more negative than stock market reactions of a country with relatively less FDI in the country. This would be due to a potential destruction of production facilities invested in, which diminishes return on investments. According to World Bank data, there was nearly no FDI before the 2003 war in Iraq. This could be due to the fact that there was an embargo imposed on Iraq at that time. Therefore, I do not take FDI into account in the analysis.
14
3.4 Methodology
The dependent variables; ‘stock market indices’, ‘government bond indices’, ‘currency exchange rates’, and ‘defense stocks’ are regressed on an interaction consisting of four parts; (i) the coefficient, (ii) the Saddam Security, (iii) a war dummy variable and (iv) a variable measuring ‘exports’, ‘imports’, ‘political relations’, ‘geographical distance’ or a dummy variable for being an ‘invading’ country or not. Commodities are also regressed against the Saddam Security interacting with a war dummy and a coefficient, but without an interaction variable for exports or imports. This is the case since only one gold index and one oil price are used which are not based in a specific country. The general formula that is used for the analysis of the stock market indices, defense industry stocks, currencies, and government bond indices, is as follows:
𝛥 ln(𝑋𝑖𝑗) = 𝛽1𝑗× 𝑆̃ × 𝑊𝑖 𝑖× 𝐸𝑦 + 𝛽2𝑗× 𝑆̃ × 𝑊𝑖 𝑖× 𝐼𝑦+ 𝛽3𝑗 × 𝑆̃ × 𝑊𝑖 𝑖 × 𝑃𝑗 +
𝛽4𝑗 × 𝑆̃ × 𝑊𝑖 𝑖 × 𝐴𝑗+ 𝛽5𝑗 × 𝑆̃ × 𝑊𝑖 𝑖 × 𝐷𝑗+ 𝛽6𝑗× 𝑆̃ × 𝑊𝑖 𝑖 + 𝛽7𝑗× 𝑆̃𝑖 (2) In which ‘Xij’ is the dependent variable for a stock market index price, defense stock price, currency price or government bond index price in time ‘i’, for country or company ‘j’. The delta depicts the first differences of the logarithm of ‘Xij’.
The other side of the equation consists of four interactions with four parts each. The ‘βxj’s are
the coefficients of the interactions for index ‘j’ and coefficient ‘x’. 𝑆̃𝑖 in all interactions stands for the first differences of the Saddam Security relative to the price of the Saddam Security of the last day, as computed in formula (1). Wi is a dummy variable for war or no war. The last part of all
interactions are the potential transmission channels, which are all stationary variables differing per country.
The transmission channels are depicted by Ey, Iy, Pj, Ay, and Dj. Ey denotes exports from Iraq
to country ‘y’ relative to the GDP of country ‘y’. I expect β1j to be positive during war, since
countries importing oil, will face less restraints with respect to the trade embargo if Saddam was removed from power. The effect before the war is expected to be negative, because uncertainty regarding oil imports could lead to a negative stock market reaction. The variable Iy stands for
imports from country ‘y’ to Iraq, relative to GDP of the country ‘y’. The expected sign for β2j is
negative, both before and during the war, as a change of regime will decrease opportunities to export to Iraq. Pj is a variable for the political relations of country ‘y’ with Iraq. The expected sign of β3j is
15
countries were his regime to lose the war. Thus asset markets of politically close countries will react negatively to attempts to remove a government favoring businesses of those countries. Ay is a dummy
variable that takes the value 1 if country ‘y’ is part of the invading nations, and 0 otherwise. The effect of being part of the invading nations is expected to be positive only during the invasion, then
β4j is expected to be positive. If the invasion is successful, it is not unthinkable that firms from those
nations will be ‘rewarded’ for their nation participating in the invasion. Dj is a variable for distance
and measures the natural logarithm of the geographical distance in kilometers. The effect of this is expected to be negative, as closer markets would appreciate it more if the conflict is over soon.
The analysis uses country fixed effects, to account for country specific effects. Using the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) it is concluded that not including lags is optimal for all dependent variables. The constant is excluded from the analysis, because I analyze first differences and do not expect a trend.
IV. Empirical Results
4.1 Output
The answer to the question if there exist trade transmission channels for the effects of war or the threat of war on stock market indices is shown in table 1. Using the stock markets output it is clear that a number of transmission channels have significant effects. For the other asset classes; government bonds, currency exchange rates, and the defense industry, effects are less clear. The results of the other asset classes are shown in table 7, appendix F. The Saddam Security variable depicts the first differences of the natural logarithm of the Saddam Security.
16
Table 1. Main regression
Variables Stock Market Indices
Saddam Security 6.345***
(1.189)
Saddam Security threat of war -6.341***
(1.260)
Saddam Security * Exports transmission channel threat of war -26.07
(71.61)
Saddam Security * Exports transmission channel during war 447.8**
(213.3)
Saddam Security * Imports transmission channel threat of war 121.8
(98.25)
Saddam Security * Imports transmission channel during war -1,495***
(280.1)
Saddam Security * Politics transmission channel threat of war -0.0668
(0.145)
Saddam Security * Politics transmission channel during war -0.790***
(0.305)
Saddam Security * threat of war, not Invading -0.0408
(0.0513)
Saddam Security * during war, not Invading 0.435***
(0.138)
Saddam Security * distance threat of war -0.00321
(0.0482)
Saddam Security * distance during war -0.662***
(0.140) All interaction variables were included separately, but none of these
did have a significant effect.
-
Observations 612
R-squared 0.110
Adjusted R-squared 0.0819
Note: Robust standard errors in parentheses. Significance levels are depicted as; *** p<0.01, ** p<0.05, * p<0.1. The
Saddam Security measures the effect of the threat of war before the war, and the effect of war during the war on the first differences of stock market indices. The Saddam Security before the war is the corrective effect of not having a war, since there is a difference in what the Saddam Security measures before and during the war. The Saddam Securities times the transmission channels measure the corrective effect of the exports, imports, political relations, invading and
geographical distance transmission channels on the general effect of the Saddam Security, broken down into transmitting the threat of war and the effect of the war itself. Country fixed effects are used in the regression.
Findings regarding the effect of war on commodities are shown in table 5 in appendix F. Transmission channels are not used in this analysis, as Dubai Crude oil is only sold in Dubai, and the gold price is also expected to be approximately the same across the world. For the oil analysis, I include a control variable for the unrest in Venezuela10. The output in table 1 clearly shows that there are significant effects of transmission channels on stock market indices, the next section will analyze these effects.
17
4.2 Discussion
The findings in table 1 are split up in a general effect of the Saddam Security, the additional effect of war, plus the effects of the exports and imports transmission channels and the effect of three control transmission channels; political relations, distance, and invading. The effect of each transmission channel is split up in an effect before and during the war. I will now explain the results one by one per transmission channel.
The base effect is the effect of the Saddam Security; it has a strong positive impact on stock market indices. However, the interaction between the Saddam Security and the war dummy gives a strong negative effect of the Saddam Security before the invasion. This means that there is a strong positive effect of the US and allies winning the war once it broke out and a strong positive plus a strong negative effect of the likelihood of war increasing.
The results indicate that there is only proof of the transmission channels working during the war. The H0 hypothesis of no effect of transmission channels is therefore rejected. The level of exports transmitting the effects of the Iraq invasion to stock markets has a positive effect. This could be due to the reason explained before; importers of Iraqi oil saw future oil imports become more accessible if the US would win the war.
The transmission channel of Iraqi imports on the other hand, has a strong negative effect during the war. The negative effect of the imports transmission channel during the war is as expected; as the war progressed in favor of the US-led coalition, investors saw future profitability decrease of firms from nations exporting relatively much to Iraq, as an export market disappeared.
As expected, the political relations transmission channel does have a strong negative effect during the war. Stock market participants were afraid that the Iraq invasion would hurt profitability of firms from nations relatively friendly with Iraq. Meanwhile, if a country not participated in the invasion, this proved beneficial for home stock market indices if the US-led coalition made progress. This negative effect is not as expected. Furthermore, and in line with my expectations, if the coalition made progress, and chances of Saddam remaining in power decreased, this decreased the positive effect of the Saddam Security for countries further away.
18
exports and the imports transmission channels at the same time, as the smaller dataset for those two dependent variables causes collinearity issues. Variance explained for the defense industry is relatively low; the adjusted R-squared is 2.02%. There is a positive effect of war on defense industry stocks, and a weak effect via the exports, imports, and control transmission channels. War affected government bonds negatively, while not being part of the invading nations and more geographical distance had a significant positive effect before the war. Variance explained with government bonds as a dependent variable is 8.75%. Variance explained with currency exchange rates as a dependent variable is higher; 12.8%. The exports transmission channel does have a significant positive effect on the price of the US dollar relative to other currencies during the war. This means that if a country’s imports from Iraq were higher before the war, that country’s currency lost value relative to the dollar if the war went in favor for the US-led coalition forces. This significant effect of currency exchange rates could be partially explained by the effect of war on government bonds and the phenomenon interest rate parity. As interest rates of certain government bonds increased relative to other government bonds, currency exchange rates needed to adjust to avoid arbitrage opportunities.
Regarding commodities, both the gold index and oil prices do not seem to be significantly influenced by the course of war. That could be because I use Dubai Crude Oil instead of the West Texas Intermediate spot oil or future prices Zitzewitz and Wolfers (2009) use. Another explanation for the absence of an effect of war likelihood and war progress on Dubai crude oil prices could be the scale of Iraqi oil production. According to the U.S. Energy Information Agency, Iraqi oil production equaled 2.04 million barrels a day in 2002. This was comparable to 2.6% of world oil production at that time.
4.3 Changes in Iraqi imports and exports following the Iraq invasion
19
Figure 3: the effect of the Iraq invasion on Iraq exports
Note: Exports from Iraq to seven other nations. The natural logarithm of the value of exports is measured per year from
1998 until 2005.
Figure 4: the effect of the Iraq invasion on Iraq imports
Note: Iraq imports from seven nations. The natural logarithm of the value of imports is measured per year from 1998
until 2005. 16 17 18 19 20 21 natural logarithm logarithm of Iraq imports in in US$ 1998 2000 2002 2004 2006 Year
China, P.R.: Mainland France
Germany Japan
Russian Federation United Kingdom
United States
Iraq Imports
0 5 10 15 20 25 natural logarithm of Iraq exports in US$ 1998 2000 2002 2004 2006 YearChina, P.R.: Mainland France
Germany Japan
Russian Federation United Kingdom
United States
20
To see if Iraq’s imports and exports changed following the Iraq invasion, and if the trends expected by stock market participants occurred after the war, graph 3 and 4 are depicted below. The graphs contain the natural logarithm of Iraq exports and imports respectively, from 1998 until 2005. The invasion took place in the year 2003. As seen in graph 3, the natural logarithm of exports from Iraq to the seven selected countries do not appear to change much after the war. Besides changes during 2003, only exports from Iraq to the US and Russia increased slightly. A robustness issue (discussed in section 4.4) regarding the exports transmission channel might explain the absence of the expected effects.
In contrary to exports, imports seem to change after the war. Imports from the US were, relative to other nations, the smallest before the war, but after the war the US tops the list. Imports from the United Kingdom also increase and imports from Germany, France and China seem to be unaffected after 2 years. Russia’s position in imports drops from 3rd before the war to 7th after the war. Thus, nations exporting more to Iraq pre-war saw exports after the war decline relative to nations exporting relatively less to Iraq pre-war. This means that firms from high pre-war exporting nations saw a market to sell goods decline relative to firms from low pre-war exporting nations, which might affect their firm value negatively.
4.4 Robustness
In this section I present the output of a number of robustness tests. First, to test if the results are also robust when several variables are excluded, I test the effects of the transmission channels by adding the transmission channels one by one for my main analysis. The robustness tests for the effect on stock market indices are shown in table 5 and 6. The same robustness tests are done for the other asset markets, and are given in appendix G, tables 9 to 14.
The robustness test for stock market indices in table 5 confirms that the control transmission channels of ‘political relations’, ‘distance’, and ‘invading’ are largely robust. All these transmission channels increase the adjusted R-squared. Table 6, in which the transmission channels are added separately confirms these findings; all three control variables increase the adjusted R-squared, and the signs of the coefficients do not change.
21
transmission channel only decreases the adjusted R-squared in both table 2 and 3. Except for the exports transmission channel, the findings regarding stock market indices are therefore largely robust.
With respect to the defense industry, government bonds, and currency exchange rates, the results are less robust. The defense industry’s adjusted R-squared only increases slightly when adding the control transmission channels ‘invading’, ‘political relations’, and ‘distance’ in table 10. Adding the exports and imports transmission channels to the regression does not increase variance explained. Variance explained by the government bonds regression only increases when adding the ‘political relations’ transmission channel in table 12, while the other control transmission channels and the transmission channels of exports and imports decrease the adjusted R-squared. Robustness for currencies as a dependent variable is higher. All main and control transmission channels separately increase the adjusted R-squared in table 13. Signs and significance levels are largely consistent. Only the ‘invading’ control transmission channel decreases R-squared in table 14.
Second, I also check for multicollinearity between independent and dependent variables. Multicorrelation between independent variables is checked for in the correlation matrix depicted in table 2. Only distance and imports, and political relations and invading are highly negatively correlated, respectively -0.7 and -0.67. The other independent variables show no high correlation with other independent variables. In table 3, the correlation between the dependent variables is checked for. It seems that only stock market indices and government bonds have an absolute correlation greater than 0.6. The Variance Inflation Factor (VIF) following a normal regression is relatively high for some interactions, but this is probably due to the fact that interactions are used in the analysis. Regressing the relevant variables apart (not in an interaction) gives VIF estimates lower than 3.
22
Table 2: Correlation independent variables
Variables Stock
Market Indices
Saddam Security
Exports Imports Political relations
Invading War dummy
Distance
Stock Market Indices 1.0000
Saddam Security -0.1132 1.0000 Exports 0.0031 -0.0001 1.0000 Imports 0.0235 -0.0010 -0.0731 1.0000 Politics -0.0049 -0.0004 -0.3191 0.3313 1.0000 Invading 0.0123 -0.0000 0.0740 -0.5273 -0.6661 1.0000 War dummy 0.0439 0.0143 0.0015 -0.0031 -0.0036 0.0025 1.0000 Distance -0.0151 0.0014 0.2937 -0.7018 -0.1388 0.3369 0.0037 1.0000
Table 3: Correlation of dependent variables
Variables Government bonds Currencies Defense Industry Stock Market Indices
Government bonds 1.0000
Currencies -0.4037 1.0000
Defense Industry -0.1261 0.2084 1.0000
Stock Market Indices -0.6437 0.4213 0.1978 1.0000
Table 4: Reversed dependency
(1)
Variables Saddam Security
Exports 1.931
(11.37)
Imports 2.974
(13.79)
Invading during war 0.00164
(0.00488)
Invading before war 0.00272
(0.00616)
Political relations 0.00358
(0.00679)
Stock market indices before war -0.703***
(0.247)
Stock market indices during war 0.931***
(0.117)
Observations 612
R-squared 0.066
23
Table 5: 1st Robustness test for stock market indices
(1) (2) (3) (4) (5) (6) (7)
Variables Stock market
indices Stock market indices Stock market indices Stock market indices Stock market indices Stock market indices Stock market indices
Saddam Security 0.345*** 0.673*** 0.762*** 2.368** 2.559*** 5.490*** 6.345***
(0.0572) (0.168) (0.150) (0.946) (0.962) (1.113) (1.189)
Saddam Security before the war -0.420*** -0.701*** -0.793*** -2.060** -2.250** -5.439*** -6.341***
(0.0598) (0.175) (0.160) (0.991) (1.006) (1.187) (1.260)
Saddam Security * Politics transmission channel before war -0.0797 -0.0544 -0.0403 -0.0391 -0.0518 -0.0668
(0.0922) (0.138) (0.131) (0.150) (0.131) (0.145)
Saddam Security * Politics transmission channel during war -0.558** -1.202*** -1.136*** -1.034*** -1.063*** -0.790***
(0.271) (0.313) (0.295) (0.357) (0.275) (0.305)
Saddam Security * before war, not Invading -0.0160 -0.0323 -0.0326 -0.0434 -0.0408
(0.0454) (0.0480) (0.0522) (0.0485) (0.0513)
Saddam Security * during war, not Invading 0.410*** 0.325** 0.291* 0.502*** 0.435***
(0.138) (0.128) (0.151) (0.131) (0.138)
Saddam Security * ln(distance) before war -0.0394 -0.0397 -0.00985 -0.00321
(0.0327) (0.0344) (0.0457) (0.0482)
Saddam Security * ln(distance) during war -0.187* -0.217* -0.543*** -0.662***
(0.108) (0.114) (0.128) (0.140)
Saddam Security * Exports transmission channel before war 2.132 -26.07
(80.01) (71.61)
Saddam Security * Exports transmission channel during war 172.4 447.8**
(259.0) (213.3)
Saddam Security * Imports transmission channel before war 111.0 121.8
(102.3) (98.25)
Saddam Security * Imports transmission channel during war -1,342*** -1,495***
(283.0) (280.1) All relevant interaction variables were included separately, but none of these did have a significant effect. - - - - -
Observations 612 612 612 612 612 612 612
R-squared 0.084 0.089 0.094 0.098 0.098 0.109 0.110
Adjusted R-squared 0.0699 0.0719 0.0746 0.0755 0.0727 0.0831 0.0819
24
Table 6: 2nd Robustness test for stock market indices
(1) (2) (3) (4) (5) (6)
Variables Stock market
indices Stock market indices Stock market indices Stock market indices Stock market indices Stock market indices Saddam Security 0.345*** 0.298*** 0.396*** 0.676*** 0.301*** 2.293** (0.0572) (0.0750) (0.0720) (0.170) (0.0664) (0.924)
Saddam Security before the war -0.420*** -0.0748*** -0.0873*** -0.0283 -0.0525** 0.158
(0.0598) (0.0228) (0.0211) (0.0508) (0.0214) (0.275) Saddam Security * Exports transmission channel before war -3.501
(62.74) Saddam Security * Exports transmission channel during war 289.4
(240.2)
Saddam Security * Imports transmission channel before war 64.97
(58.93)
Saddam Security * Imports transmission channel during war -244.1
(173.2)
Saddam Security * Politics transmission channel before war -0.0797
(0.0923)
Saddam Security * Politics transmission channel during war -0.558**
(0.273)
Saddam Security * before war, not Invading -0.0310
(0.0302)
Saddam Security * during war, not Invading 0.0668
(0.100)
Saddam Security * ln(distance) before war -0.0274
(0.0324)
Saddam Security * ln(distance) during war -0.229**
(0.108) All relevant interaction variables were included separately, but none of these did have a significant effect. - - - -
Observations 612 612 612 612 612 612
R-squared 0.084 0.084 0.085 0.088 0.085 0.088
Adjusted R-squared 0.0699 0.0692 0.0703 0.0733 0.0700 0.0724
25
V. Conclusion and recommendations
To conclude, a number of asset classes was influenced by the Iraq invasion of 2003 through trade transmission channels. The main asset class researched in this thesis, stock market indices, are positively affected by coalition progress in the invasion of Iraq in 2003 through the exports transmission channel and negatively by the imports transmission channel. Using a number of control transmission channels, I find that these results are robust for the imports transmission channel but not so much for the exports transmission channel. No proof is found for the pre-war effects of the likelihood of armed conflict on stock market indices.
Stock market participants seem to expect that firms from countries with relatively high pre-war exports to Iraq would lose market share to firms from countries with relatively low pre-pre-war exports to Iraq. Effects of the Iraqi export transmission channel were positive, however; results regarding this transmission channel were not robust. A short analysis of changes in Iraqi international trade seems to confirm the expectations of stock market participants, although further research is needed to conclude if these changes in exports and imports are really significant.
Based on my findings, I would recommend policy makers to take the trade transmission channels into account when determining the financial and economic consequences of armed conflicts. Exports to and imports from a nation experiencing armed conflict could be hurt, or new opportunities could arise depending on the armed conflict outcome. This could in turn affect the profitability of businesses of a country. From a finance perspective, it is important for asset markets participants to understand what transmission channels determine the effects of armed conflict on asset markets. Knowing how the effects of armed conflict on asset markets are moderated could help with optimizing an investment strategy.
26
the 2003 invasion. Also, political relations are difficult to measure, making this a weak spot in this research. Recommended is that future research using more sophisticated measures of political relations between countries, more stock market indices and more conflicts in general should be performed. How different transmission channels affect different industries within economies is also not touched upon in this thesis, and would be an interesting subject for future research as well. Another important topic for future research in the field of armed conflict are prediction markets. Prediction markets proved to be vital for this thesis. It would therefore be of utmost importance for future research in this field that prediction markets regarding armed conflicts subsist to ensure future availability of data for armed conflict research.
Appendices
Appendix A: Stock market indices
Index Index number Observations
Russia: RUSSIA RTS INDEX (RSRTSIN) 1 90
US: S&P 500 COMPOSITE (S&PCOMP) 2 87
UK: FTSE 100 (FTSE100) 3 90
France: FRANCE CAC 40 (FRCAC40) 4 90
Germany: DAX 30 PERFORMANCE
(DAXINDX) 5 89
Japan: NIKKEI 225 STOCK AVERAGE (JAPDOWA)
6 83
China: SHANGHAI SE B SHARE (CHSBSHR)
7 90
Appendix B: Exchange rates relative to the United States dollar
Currency Index number Observations
Russian Ruble 1 90
Euro 2 90
United Kingdom Pound 3 90
Japanese Yen 4 90
Chinese Renminbi 5 90
Appendix C: Government bond indices
Country Index number Observations
United Kingdom 1 90
United States 2 90
France 3 90
Japan 4 90
27
Appendix D: Defense companies
Company Index number Country Observations
Raytheon 1 US 90 Smith Wesson 2 US 90 Boeing 3 US 90 General Dynamics 4 US 90 Lockheed Martin 5 US 90 Norinco 6 China 90 AVIC 7 China 90 Daikin 8 Japan 90 Rheinmetall 9 Germany 90 Thales 10 France 90 British AE 11 UK 90 Cobham 12 UK 90 Babcock 13 UK 90 Rolls Roys 14 UK 90 Howa 15 Japan 90
Appendix E: Similarity in UN voting patterns between Iraq and seven other countries Country Country code Similarity in voting pattern
Iraq 645 1 US 2 .26316 UK 200 .50877 Russia 365 .60526 China 710 .96364 France 220 .55455 Germany 255 .58772 Japan 740 .64035
Appendix F: additional results
Table 7: Effect of war on commodities
(1) (2)
Variables Gold index Oil index
Saddam Security 0.00454 0.000400
(0.00314) (0.00173)
War dummy 0.000823 -0.00355
(0.00784) (0.0119)
Saddam Security during war -0.00327 -0.00632
(0.00349) (0.00524) Venezuela crisis -0.452 (0.382) Constant -0.000898 -0.000786 (0.00393) (0.00348) Observations 89 89 R-squared 0.152 0.066
28
Table 8: the effect of transmission channels on Defense Industry stocks, Government bonds and currencies
(1) (2) (3)
Variables Defense Industry
Stock prices Government Bonds Indices Currency exchange rates relative to US $ Saddam Security 28.48** -0.730*** 0.0413 (11.98) (0.223) (0.132)
Saddam Security before the war -30.24** 0.798*** -0.0269
(12.95) (0.241) (0.0287)
Saddam Security * Exports transmission channel before war -132.3 1.833 -160.9
(2,414) (27.28) (118.2)
Saddam Security * Exports transmission channel during war 11,965* -4.142 383.5**
(6,592) (47.64) (173.2)
Saddam Security * Imports transmission channel before war 100.4 - -
(5,802)
Saddam Security * Imports transmission channel during war -27,744* - -
(15,459)
Saddam Security * Politics transmission channel before war -0.0491 -0.0283 0.00426
(2.625) (0.0769) (0.0525)
Saddam Security * Politics transmission channel during war 11.50 0.279 -0.333**
(7.120) (0.181) (0.140)
Saddam Security * No war, not Invading 0.116 0.794*** -0.797*
(0.144) (0.245) (0.414)
Saddam Security * war, not Invading -0.122 -0.0138 -0.772**
(0.390) (0.0346) (0.378)
Saddam Security * ln(distance) before war 0.197 -0.00388 -0.00522
(0.705) (0.00922) (0.0208)
Saddam Security * ln(distance) during war -3.932** 0.0592** 0.120**
(1.785) (0.0238) (0.0557)
Observations 1,335 445 445
R-squared 0.033 0.118 0.158
Adjusted R-squared 0.0202 0.0875 0.128
29
Appendix G: Robustness checks
Table 9 & 10: Defense industry robustness tests Table 9: Defense Industry robustness test
(1) (2) (3) (4) (5) (6) Variables Defense industry stock prices Defense industry stock prices Defense industry stock prices Defense industry stock prices Defense industry stock prices Defense industry stock prices Saddam Security 0.208*** 0.233** 0.191** 0.270** 0.199*** 3.718*** (0.0651) (0.108) (0.0793) (0.118) (0.0742) (1.414)
Saddam Security before the war -0.235*** -0.0429 -0.0308 -0.0781 0.0298 -1.261
(0.0763) (0.0628) (0.0539) (0.0916) (0.0458) (0.855)
Saddam Security * Exports transmission channel before war 79.09
(205.3)
Saddam Security * Exports transmission channel during war -108.7
(345.4)
Saddam Security * Imports transmission channel before war 50.28
(293.9)
Saddam Security * Imports transmission channel during war 266.5
(608.9)
Saddam Security * Politics transmission channel before war 0.0998
(0.141)
Saddam Security * Politics transmission channel during war -0.113
(0.220)
Saddam Security * before war, not Invading 0.0946
(0.0736)
Saddam Security * during war, not Invading 0.0310
(0.139)
Saddam Security * ln(distance) before war 0.141
(0.0976)
Saddam Security * ln(distance) during war -0.401**
(0.160) All interaction variables were included separately, but none of these did have a significant effect. - - - -
Observations 1,335 1,335 1,335 1,335 1,335 1,335
R-squared 0.009 0.010 0.009 0.010 0.015 0.020
Adjusted R-squared 0.00303 0.00290 0.00235 0.00369 0.00825 0.0136
30
Table 10: Defense Industry robustness test
Note: see note to table 1.
(1) (2) (3) (4) (5) (6) (7)
Variables Defense stock
prices Defense stock prices stock prices Defense stock prices Defense stock prices Defense stock prices Defense stock prices Defense
Saddam Security 0.208*** 0.266** 0.519 5.589*** 5.617*** 5.640*** 28.48**
(0.0651) (0.118) (0.357) (1.694) (1.707) (1.788) (11.98)
Saddam Security before the war -0.235*** -0.344** -0.385 -7.183*** -7.289*** -7.143*** -30.24**
(0.0763) (0.150) (0.408) (1.874) (1.875) (1.993) (12.95)
Saddam Security * Politics transmission channel before war 0.0998 -0.143 0.0536 -0.00374 0.0895 -0.0491
(0.141) (0.248) (0.187) (0.223) (0.208) (2.625)
Saddam Security * Politics transmission channel during war -0.113 -0.403 -1.056** -1.036* -1.036** 11.50
(0.221) (0.418) (0.417) (0.580) (0.425) (7.120)
Saddam Security * before war, not Invading - 0.0972 0.115 0.111 0.116
- (0.101) (0.111) (0.108) (0.144)
Saddam Security * during war, not Invading - 0.332 0.326 0.340 -0.122
- (0.231) (0.284) (0.271) (0.390)
Saddam Security * No war, invading -0.145 - - - -
(0.126)
Saddam Security * war, Invading -0.173 - - - -
(0.241)
Saddam Security * ln(distance) before war 0.172* 0.185** 0.160 0.197
(0.0900) (0.0875) (0.0997) (0.705)
Saddam Security * ln(distance) during war -0.569*** -0.573*** -0.575*** -3.932**
(0.183) (0.192) (0.194) (1.785)
Saddam Security * Exports transmission channel before war -89.11 -132.3
(192.1) (2,414)
Saddam Security * Exports transmission channel during war 30.51 11,965*
(479.1) (6,592)
Saddam Security * Imports transmission channel before war -205.1 100.4
(462.9) (5,802)
Saddam Security * Imports transmission channel during war -114.1 -27,744*
(1,125) (15,459) All interaction variables were included separately, but none of these did have a significant effect. - - - - -
Observations 1,335 1,335 1,335 1,335 1,335 1,335 1,335
R-squared 0.009 0.010 0.017 0.031 0.032 0.032 0.033
31
Table 11 & 12: Government bonds indices robustness test Table 11: Government bonds indices robustness test
(1) (2) (3) (4) (5) (6) Variables Government bond indices Government bond indices Government bond indices Government bond indices Government bond indices Government bond indices Saddam Security -0.0874*** -0.0773*** -0.0824*** -0.165** -0.0759*** -0.276 (0.0115) (0.0125) (0.0177) (0.0675) (0.0119) (0.269)
Saddam Security before the war 0.106*** 0.0170*** 0.0196*** 0.0370* 0.0146*** 0.000963
(0.0124) (0.00589) (0.00635) (0.0197) (0.00535) (0.0985)
Saddam Security * Exports transmission channel before war 6.388
(20.42)
Saddam Security * Exports transmission channel during war -48.91
(45.14)
Saddam Security * Imports transmission channel before war -12.91
(47.73)
Saddam Security * Imports transmission channel during war -54.75
(99.71)
Saddam Security * Politics transmission channel before war -0.0365
(0.0353)
Saddam Security * Politics transmission channel during war 0.150
(0.120)
Saddam Security * before war, Invading 0.00935
(0.00947)
Saddam Security * during war, Invading -0.0294
(0.0249)
Saddam Security * ln(distance) before war 0.00202
(0.0114)
Saddam Security * ln(distance) during war 0.0219
(0.0320) All interaction variables were included separately, but none of these did have
a significant effect. - - - -
Observations 445 445 445 445 445 445
R-squared 0.108 0.110 0.109 0.115 0.113 0.109
Adjusted R-squared 0.0941 0.0932 0.0925 0.0984 0.0970 0.0927
32
Table 12: Government bonds indices robustness test
(1) (2) (3) (4) (5) (6) Variables Government bond indices Government bond indices Government bond indices Government bond indices Government bond indices Government bond indices Saddam Security -0.0874*** -0.164** -0.169** -0.736*** -0.730*** -0.730*** (0.0115) (0.0678) (0.0793) (0.209) (0.223) (0.223)
Saddam Security before the war 0.106*** 0.201*** 0.202** 0.808*** 0.798*** 0.798***
(0.0124) (0.0704) (0.0829) (0.224) (0.241) (0.241) Saddam Security * Politics transmission channel before war -0.0365 -0.0235 -0.0317 -0.0283 -0.0283 (0.0353) (0.0595) (0.0567) (0.0769) (0.0769)
Saddam Security * Politics transmission channel during war 0.150 0.165 0.287* 0.279 0.279
(0.120) (0.165) (0.154) (0.181) (0.181)
Saddam Security * before war, not Invading 0.197** 0.804*** 0.794*** 0.794***
(0.0871) (0.226) (0.245) (0.245)
Saddam Security * during war, not Invading -0.00508 -0.0154 -0.0138 -0.0138
(0.0296) (0.0287) (0.0346) (0.0346)
Saddam Security * ln(distance) before war -0.00402 -0.00388 -0.00388
(0.00896) (0.00922) (0.00922)
Saddam Security * ln(distance) during war 0.0595** 0.0592** 0.0592**
(0.0235) (0.0238) (0.0238)
Saddam Security * Exports transmission channel before war 1.833 1.833
(27.28) (27.28)
Saddam Security * Exports transmission channel during war -4.142 -4.142
(47.64) (47.64) All interaction variables were included separately, but none of these did
have a significant effect.
- - - -
Observations 445 445 445 445 445 445
R-squared 0.108 0.115 0.115 0.118 0.118 0.118
Adjusted R-squared 0.0941 0.0964 0.0925 0.0917 0.0875 0.0875
33
Table 13 & 14: Currency robustness test Table 13: currency robustness test
(1) (2) (3) (4) (5) (6) Variables Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$
Saddam security before war - -0.00743 -0.0211* -0.0584 -0.0275 -0.0379
(0.0103) (0.0127) (0.0360) (0.0244) (0.167)
Saddam Security during war 0.105*** 0.0491*** 0.104*** 0.240*** 0.0890*** -0.405
(0.0230) (0.0165) (0.0253) (0.0480) (0.0237) (0.381)
Saddam Security * Exports transmission channel before war -131.9
(109.9) Saddam Security * Exports transmission channel during war 500.5***
(143.0)
Saddam Security * Imports transmission channel before war 20.90
(23.61)
Saddam Security * Imports transmission channel during war -94.82*
(53.18)
Saddam Security * Politics transmission channel before war 0.0622
(0.0396)
Saddam Security * Politics transmission channel during war -0.232***
(0.0495)
Saddam Security * before war, not Invading 0.0128
(0.0274)
Saddam Security * during war, Invading -0.0129
(0.0333)
Saddam Security * ln(distance) before war 0.00244
(0.0193)
Saddam Security * ln(distance) during war 0.0583
(0.0460) All interaction variables were included separately, but none of these
did have a significant effect. - - - -
Observations 445 445 445 445 445 445
R-squared 0.084 0.131 0.089 0.103 0.088 0.087
Adjusted R-squared 0.0695 0.115 0.0727 0.0870 0.0709 0.0704
34
Table 14: Currency robustness test.
(1) (2) (3) (4) (5) (6) Variables Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Currency exchange rates with US$ Saddam Security -0.0173 -0.0584 -0.0584 0.0215 0.0413 -0.818 (0.0113) (0.0360) (0.0360) (0.151) (0.132) (0.714)
Saddam Security during the war 0.105*** 0.299*** 0.374*** -0.705* -0.772** 2.134
(0.0230) (0.0608) (0.0842) (0.420) (0.378) (1.373)
Saddam Security * Politics transmission channel before war 0.0622 0.0606 0.0746 0.00426 0.132*
(0.0397) (0.0517) (0.0518) (0.0525) (0.0767)
Saddam Security * Politics transmission channel during war -0.232*** -0.323*** -0.501*** -0.333** -0.638***
(0.0496) (0.0782) (0.117) (0.140) (0.122)
Saddam Security * before war, not Invading 0.00129 -0.000139 -0.0269 0.0573
(0.0326) (0.0323) (0.0287) (0.0603)
Saddam Security * during war, not Invading -0.0745* -0.794* -0.797* 1.908
(0.0430) (0.437) (0.414) (1.325)
Saddam Security * ln(distance) before war -0.0104 -0.00522 0.0789
(0.0179) (0.0208) (0.0756)
Saddam Security * ln(distance) during war 0.133** 0.120** -0.0802
(0.0550) (0.0557) (0.131)
Saddam Security * Exports transmission channel before war -160.9
(118.2)
Saddam Security * Exports transmission channel during war 383.5**
(173.2)
Saddam Security * Imports transmission channel before war 200.4
(147.2)
Saddam Security * Imports transmission channel during war -477.6**
(215.7) All interaction variables were included separately, but none of these did have a significant
effect. - - - -
Observations 445 445 445 445 445 445
R-squared 0.084 0.104 0.107 0.121 0.158 0.158
Adjusted R-squared 0.0695 0.0851 0.0845 0.0949 0.128 0.128
