CHAPTER THREE

From Popper to Kuhn to Lijphart (1971), debates surrounding the necessity for researchers to find precise methods of induction have often included the criticism that qualitative research lacks the ability to be a reliable alternative to quantitative techniques. Now the argument has taken a new turn. Instead of trying to encourage all social scientists to use the quantitative method, peers suggest that the qualitative approach really can work, as long as it is used in a manner drawing it ever closer to the quantitative style. The APSA-CP debates this trend (Winter edition, 1996), suggesting ways to create a process for the replication of qualitative research. In addition, the degree of policy relevancy can often prescribe a certain methodological path.

The approach and methodology of qualitative research may provide for important and credible research. Some approaches produce better results than others. The importance of choosing the correct path reflects the desire to reduce error and raise reliability. Researchers hold different opinions on how to approach qualitative, quantitative and case study research. A discussion of the various approaches and methods may suggest specific actions to take when conducting qualitative research.

In the next section I will briefly summarize several important pieces of literature discussing how to approach research. I will discuss the debate over the use of the qualitative method first, then other suggestions on research methods and approaches. Then I will compare and contrast the various positions, and critique possible problem areas. Finally, I will offer suggestions for future research and areas of debate.

Gary King, Robert Keohane and Sidney Verba (1994) delineate specific ways in which to achieve the most precise qualitative research design in their book Designing Social Inquiry: Scientific Inference in Qualitative Research. King, Keohane and Verba use an argument similar to Ragin’s (1994) in their support of the qualitative research method. The authors acknowledge that when cases are few, the qualitative method is the best choice, but they also believe that a research design can incorporate both quantitative and qualitative methods. Ragin, however, sees a sharp contrast between the three styles, especially between the qualitative and quantitative methods, while King, Keohane and Verba proposed that these differences are not what they seem: “A major purpose of this book is to show that the differences between the quantitative and qualitative traditions are only stylistic and are methodologically and substantively unimportant” (4). Yet, both qualitative and quantitative methods suggest differing avenues of research. The substance of each is distinct due to different answers to different questions. As Ragin points out, the qualitative method tends to provide more insight and depth into a subject, while the quantitative method tends to show generalities. The ability of each method to uncover different types of information is what makes them work so well when put together or when used separately to get at different questions around a common theme. To generalize that both methods are substantively the same because they both utilize logical inference in the research program is too broad of a grouping.

King, Keohane and Verba do provide specific procedures to follow when attempting a qualitative research design. They suggest increasing the observations when conducting case studies or low n research. Instead of looking at one country as a case study in itself, a researcher should observe specific points in time or policy changes. Increasing the number of observations allows for hypothesis testing. Increasing n provides for a better research design.

While increasing the n does appear to facilitate greater explanatory power, making qualitative analysis adhere to mathematical equations as suggested by King, Keohane and Verba almost takes away from the beauty of qualitative analysis.[1] Almond (1990) questions the strict adherence to the scientific structure of social science in “Clouds, Clocks, and the Study of Politics.” Almond shows that while the hard sciences can easily adhere to the scientific method, just as clocks—or time—can be shown in a structured manner, social science isn’t the same type of animal. The “cloud-like” nature of social phenomena is ever changing, reshaping itself into different outlines with growing and shrinking depths and mass. In this way, the qualitative method allows the researcher to expand and contract with his or her information. Johnson (1977) suggests that while qualitative research is less systematic than is quantitative, it is sensitive to context (semantic, political, historical, and situational). Such sensitivity may provide important data not observable in quantitative research. Again, when the question being asked does not easily fit to quantitative research, instead of changing the question or neglecting the research, the qualitative method allows for inquiry.

King (1993) also suggests that both qualitative and quantitative research can be done systematically and are important in theory building. He even suggests that the quantitative method is not always used in a proper manner, resulting in less than precise information. King proposes that qualitative researchers avoid being normative and do research that builds, or can be

built upon, especially through replication. In other words, another researcher should be able to repeat exactly the same process as was done the first in order to test further the theory. Besides increasing the n of observations, there should be rules that researchers must follow in order to produce good, systematically obtained results, which could be replicated by someone else.

Alexander George (1982) also supports the use of a rigorous qualitative research approach. While some theoreticians may feel that the comparative method is not as good as experimental and statistical methods, it can be made more systematic and reliable. George also suggests raising the number of cases that are to be observed. In addition, combining similar variables reduces the problem of more variables than cases. George supports King’s call for more rigorous rules in qualitative research. George states that researchers should view their theoretical questions from either the agreement or difference approach. The agreement approach considers similar independent variables that lead to similar outcomes. The difference approach relies on similar independent variables which lead to different outcomes. In this way, an analysis of possible reasons for the agreement or difference between dependent variables provides useable information for theory building.

On the other hand, there are academics who do not believe there is a problem with the normative aspect of some qualitative research. Rosenau (1987) contends that case studies and area studies are important without changing them. Case studies can explain events within a country better than can comparative quantitative methods, leading to better international or foreign policy theory. The question being asked may only be answerable through an in-depth analysis of a single case viewed through the changing environment of a state in flux. If one is viewing revolution in the United States, there is only one case, which may call for a comprehensive analysis of the event. Additionally, the historical content of case studies is important to the contextual aspect of most qualitative approaches and to theory building in general. Contextual information may not easily fit into a set of rules that may limit the intuitive nature of qualitative research.

On another note, George (1982) includes policy relevance as an important issue that must be included when conducting research. He believes that comparing cases has policy relevance. More important, however, is the researcher’s ability to be policy specific—how does the research help policy makers make their decisions? In order to be more policy relevant, researchers should have precise meanings for concepts and terms, not be too parsimonious, and use language understandable to policy makers. Furthermore, policy makers often find academic information problematic. They believe that academic research is not as unbiased as the quantified techniques imply. Data-based information distorts the reality of political situations. Therefore, policy makers contend that “statecraft is an art not science.” George also notes the grumbling of researchers towards policy makers and how they make decisions. Academics believe policy makers use “warmed over” realism, or apply theories improperly, and then blame researchers when theories fail. Policy makers also tend to look more at historical events than at the specifics of an event or possible forces other than power. Plus, policy makers always have political considerations that affect policy making in ways researchers believe it shouldn’t. A better understanding of what policy makers need from research allows researchers to focus parts of their analysis towards these needs.

While King, Keohane and Verba present a variety of helpful suggestions on how to advance a more scientific research design, some of their analysis appears to call for a limit on those characteristics that add to the qualitative method’s uniquely open and intuitive nature. Nevertheless, the insight they provide into good methods and approach is useful for all researchers. If a large segment of political scientists takes its suggestions under consideration, political science as a whole will be better off. Utilizing systematic models of research and logical inference can help make qualitative research more credible.

On the other hand, applying statistical structures to qualitative research may be going too far. King, Keohane, and Verba’s use of numeric representation may be unnecessary and meaningless. By trying to represent numerically low n analysis, important contextual information is lost, and the numbers may not necessarily interpret the event. Applying numbers to contextual information not only takes away from the unique nature of qualitative research but also pretends to quantify an analysis on data collected qualitatively. It could very well be a cosmetic fix to make qualitative research appear quantified. In addition, some researchers may argue that increasing points of observations within a case is simply a superficial process and does not really create more reliable data. Yet, King, Keohane, and Verba do provide food for thought on good social science research and on how theory-building research can be done.

In his Presidency article, King does well to caution both qualitative and quantitative researchers to be more rigorous in their research. Especially useful is his suggestion that a set of rules or structures should be developed to make qualitative and case study research more replicable. Rules and structures help focus researchers, creating better theory building results. Unfortunately, replication is very difficult in qualitative and case study research. The reason many researchers use these methods is because of the type of information desired or the question being asked, making the data collection process specific and difficult to replicate. In addition, while King’s caution against normative research is important, it is normative work that sometimes pushes limiting boundaries in social science. Yet, the warnings provided by our peers in the social sciences can help in the theory making process.

Social science research can benefit from the rules and structures that George (1982) suggests and added to by King (1993). The guidelines suggested by George are important for creating more reliable data and contributing to theory building. His outline of the agreement/difference approach helps make replication at least a concept to consider when setting up a research design. However, for some case studies there may be limits on the effect the rules and structures will have. Not only could the limiting aspect of rules be of little use, but they may actually hide important data that could be produced otherwise.

Conversely, Rosenau’s (1987) support of the area/case study approach reinforces the importance of contextual information. Sometimes quantitative work does not adequately explain why events occur. There are also occasions when large n data collection masks important differences, or suggests differences that are actually unimportant. Contextual data that uncover cultural specifics can be used to support or qualify quantitative data. Yet, it may still be valuable to proscribe the type of rules and structures suggested by King (1993) and George (1982). If there isn’t any systematic model for obtaining information, there may be a question of whether the information is any better than the author’s opinion. Everyone processes information differently, and one perspective may not be the correct or only view. Structures could help make the textual information more usable for other researchers.

Finally, George’s (1982) suggestion of policy relevance should be considered by all researchers. Making sure research provides useable information for those who partake in policy action links academic research to the individual in the field. In addition, asking researchers to provide specific concepts and terms can provide policy makers greater access to academic research. Even with the best approach, if concepts and terms are vague, the study becomes unreplicable and the information unusable. Clearer concepts also help policy makers understand important theoretical ideas more fully. Recognizing the importance of specific concepts can also bring research and practice together.

Researchers need to think about what they want theory work to do. Do they want to be policy specific, do theory for theory’s sake, or undertake some happy medium in between? Also, some structures or rules for obtaining information will allow for better comparability, theory-building, and the ability for research to be replicated. Focusing the researcher is important for producing usable research.

In addition, researchers should continue the discussion on methodology and approach. As political science grows, instructions on how to conduct research can only benefit social science. Researchers do not have to accept everything offered, but some aspects of prescriptive offerings may ring a bell or spark an interest, providing a focus for better research. Still, political scientists should not rule out the value of work that does not follow all the rules and structures set out by peers, but should recognize the possible limitation of the work.

How researchers approach their work as well as the methods used to obtain data are important concerns of theoreticians. By providing structure and guidelines for qualitative, quantitative and case study research, more systematic and replicable research can be produced. Researchers do not have to accept or use everything suggested, but focusing researchers on specific paths will be better for theory building than will a variety of approaches and methods that fail to connect or support other research. Continued discussion on how to approach research and what methods to use will only create a better path for future research.

Similarities and differences emerge in a comparison of Iraq and North Korea. Both Saddam Hussein and Kim Il Sung used purges to limit any possible opposition to their reign. These purges focused on those who sought to change the regime and those who appeared to be gaining popular support. There was no place for individuals more popular than the leader, so individuals who held the hearts of the people often found themselves relegated to lesser positions, removed from power, or even killed. In addition, both Hussein and Kim placed family members into positions of power. Family and kinship loyalties were stronger than party relationships. Vacancies left open through purges were filled by those closest to the leader. Furthermore, the use of the cult of personality reinforced each dictator’s hold over his people, using cultural structures to support his reign. Both Iraq and North Korea are ancient kingdoms with long histories and eras of cultural development, with leaders seeking to enter into the world community on their own terms.

Important to my analysis is that both Iraq and North Korea should be placed in the dependent of the enemy and rogue state image categories. Both states were dependent upon the Soviet Union for economic and military assistance. They accepted military advisors into their countries, participated in economic programs, and accepted military and economic aid. Extended deterrence from the Soviet Union to both Iraq and North Korea enabled Iraq and North Korea to take policy actions with the knowledge that the use of force in retaliation would be unlikely from the United States. When military assistance from the Soviet Union ended for Iraq and North Korea, there was an attempt to obtain weapons of mass destruction. More importantly, when the Soviet Union ceased to exist and the umbrella of Soviet nuclear deterrence broke down, Iraq and North Korea, understanding the necessity of nuclear capability both militarily and politically, sought to obtain nuclear weapons and weapons of mass destruction. In addition, both states have been accused of going against the desires of the international community and of taking threatening actions against their neighbors. Thus, both states fall into the category of the dependent of the enemy and rogue state image categories.

Interestingly, while some policy preferences of United States policy makers towards Iraq and North Korea are similar, in some cases they differ. While Iraq often incurs the bombing of its cities and military assets, the U.S. tends to use diplomatic measures towards North Korea. The question arises as to what creates both similar and different foreign policy actions. Are there subtle differences in held imagery of the states that suggest a similar image perception but different action? Are image perceptions similar resulting in similar response alternatives, suggesting the difference in policy preferences[2] the likelihood that North Korea has one or two nuclear weapons while Iraq only has chemical or biological weapons? Or is it simply that Iraq has a desired natural resource, oil, while North Korea has nothing of interest?

The research problem centers on the images of the dependent of the enemy and the rogue state. While the terms “dependent of the enemy” and “rogue state” are used in the literature, very few studies investigate this imagery. How policy makers perceive “rogue states” may not equate to the theoretically based image. The policy maker and the researcher may not be on the same page. For the theory to be accurate, it must reflect policy makers’ perceptions. The phenomena to which I am seeking to bring clarity are how policy makers use images when dealing with rogue

states and how the rogue state image developed out of the dependent of the enemy image. Also of importance are the similarities and differences between foreign policy preferences of policy makers towards Iraq and North Korea.

Three questions result from an analysis of the literature. First, the literature suggests there is a relationship between the dependent of the enemy and the rogue state image: what is the connection between the dependent of the enemy image and the rogue state image? Second, analysis proposes we focus on the policy responses of the perceiver state: do policy makers perceive an image of a rogue state in their policy decisions towards states that fit the rogue state image? In other words, are policy makers’ policy preferences the same as what is predicted by image theory on rogue states: is the rogue state image salient for policy makers? Third, questions also reflect the similarities and differences in foreign policy preferences towards Iraq and North Korea: why are policy preferences towards Iraq often more retributive while those towards North Korea appear more rehabilitative? It is possible that policy makers either place North Korea to the right of Iraq within the image category (farther away from the prototype but still a rogue state), or perceive North Korea to be more complex or in a different category completely.

The first hypothesis that comes from the questions reflects the relationship between the dependent of the enemy and the rogue state images: a change from the dependent of the enemy image to the rogue state image will produce a change in policy preferences towards the perceived state. The second hypothesis centers on the development of the rogue state image: the closer the held image of a state is to the prototypical rogue state image, the more policies of decision makers move towards expected rogue state policies. A third hypothesis reflects the comparability of policy makers’ policy preferences towards Iraq and North Korea: a rogue state image observed as less threatening will result in rehabilitative policy preferences while a rogue state image observed as more threatening will result in retributive policy preferences. The perception of North Korea as less threatening could relate to the successful containment of that state, even though North Korea was thought to have been further along in its nuclear development. Also, if Iraq is observed as a greater threat, this could be due to a greater opportunity to enact punishment or its threat to oil reserves throughout the region, or both.

The independent variable is the image held by the policy maker of the perceived states. Included in the analysis are both the dependent of the enemy and the rogue state images. The independent variable will be operationalized through content analysis of documents (described below) based on Cottam’s (1994) code book. Image indicators include “(1) perceptions of a country’s capability, culture, and intention; (2) event scripts, reflecting lessons from history that policy makers use to understand the behavior of a country or to predict its behavior; and (3) response alternatives that were consistently considered appropriate for use vis-a-vis a country” (188). Cottam operationalizes these variables in the appendix of Images and Intervention, pages 188-189:

c. The likelihood that the country would resort to the use of military force to achieve its goals

d. Whether the country’s military force was superior, equal, or inferior to the U.S. military force

d. The government’s ability to carry out a policy, achieve a goal, or abide by an agreement

e. Whether the decision structure was multitiered or monolithic (monolithic countries are assumed to be more capable since they do not have to please their publics, interest groups, or bureaucratic interests)

a. The capacity and stability of the country’s economy (industrial potential, agricultural self-sufficiency, growth rate, potential for growth and development)

b. The interaction between the U.S. economy and the other country’s economy (permeability of other economy; threat to or opportunity for the United States)

(includes social norms, literacy, religion, standard of living, scientific and technological capabilities, racial composition, nationalism and the public-mindedness of citizens)

a. Leader’s willingness to bargain, change tactics, and shift policy in response to U.S. initiatives

4. Whether decision structure is multitiered or monolithic (those seen as multitiered are seen as less threatening)

b. Historical incident used as lesson regarding appropriateness of techniques for dealing with conflict or issue at hand

1. Instruments deemed appropriate for use in a conflict with the country (includes military threat or actual force, economic incentives to economic sanctions, diplomatic protests, bilateral and multilateral negotiations, or simply doing nothing; those perceived as weaker are dealt with in a more coercive fashion)

2. Bargaining (those considered equal are dealt with as equals; inferiors are not bargained with)

In addition to the above factors of the independent variable, statements reflecting the reliance on the enemy state will be collected. These could be part of the military capability, part of the economic capability, and also statements suggesting a reliance on the government of the enemy. These statements reflect aspects of the dependent of the enemy image as well as a possible transition image. Once I delved into the work of analyzing the contents of the documents, I found that several sections needed to be added to the measure instrument A in order to adequately incorporate the specifics of the dependent of the enemy image. In capability, under military strength and capability, I include g). military assets obtained from a “parent” state, h). superpower conflict part of state’s military capability and i). military policy dictated by “parent” state. Within the first section of culture where the perceived state is compared culturally to the U.S., I add –and/or international culture . Under intentions, I add 6). State’s goals and policies influenced by “parent” state. Under response alternatives, I include in 1). sale or gift of military assets, add 3). past policies a). successes and b). failures, and add 4). Call to superpower to intervene in crisis.

An analysis of public and private documents pinpoints statements reflecting the above criteria, developing policy makers’ perceived images of a state. In addition, the relative level of policy makers’ perceived threat emerges through an analysis of the above image indicators, enabling the comparison between Iraq and North Korea on the relative strength of the perceived image. Image indicators collected prior to the origination of the crisis situation reflect the image held by the policy maker at the beginning of the crisis.

Policy makers perceive the dependent of the enemy state as inferior militarily, culturally and economically. Yet, the dependent of the enemy has behind it the strength of the enemy’s military. Response alternatives include diplomatic exchanges and economic sanctions. The use of force is a possibility, but is very unlikely as the perceiver state would not want to incur the wrath of the enemy’s military. Response alternative that address the enemy’s involvement in the dependent state would rank high on policy makers’ list. While the rogue state is also perceived militarily, culturally and politically inferior, it is aggressive in nature and headed by an elite leadership that follows anti-western policies and seeks weapons of mass destruction. Response alternatives towards a prototypical rogue state include compellent strategies and the use of force, economic sanctions, attempts to ruin the leadership and create instability, and restrictive treaties. The two images overlap as they are both seen as inferior militarily, culturally, politically and economically. Yet, the enemy’s military strength stands behind the dependent of the enemy image and not the rogue state. Furthermore, the rogue state adopts anti-western policies, seeks weapons of mass destruction and is generally more aggressive in nature than is the dependent of the enemy. Response alternatives also overlap in policy makers’ calls for the use of force and economic sanctions. However, policy makers use of diplomacy towards the prototypical rogue state would involve less give and take and compromise than with the dependent of the enemy and would include treaties limiting rogue states’ actions.

One of the interesting possibilities that may come out of the data centers on the deterrent capabilities of the perceived dependent of the enemy and rogue state images. Since the dependent of the enemy has the military of the enemy as part of its deterrent capabilities, this capability was lost after the end of the cold war, along with the dependent of the enemy image. It may be that some states previously in the dependent of the enemy image determined that in order to achieve the deterrent level previously held through the umbrella of the enemy, the perceived state must seek weapons of mass destruction. Thus we end up with a rogue state.

The interaction between the dependent of the enemy and the enemy will be key observing a difference between the dependent of the enemy and rogue state images. There should be some discussion of the enemy in conjunction with the dependent of the enemy. Policy makers’ statements should imply the selling of weapons to the dependent, monetary and military support, a government observed as bending to the will of the enemy, and other remarks toward the enemy-dependent relationship. The rogue state image does not have the support of the enemy. Seeking power, rogue states may look to regional opportunities or threats rather than furthering the aspirations of the enemy.

Fig. 3 Identifying the Differences in the Dependent of the Enemy and Rogue Images

The dependent variable is a range of policy preferences, or those actions that policy makers voice as appropriate towards the perceived state in a crisis situation. I operationalize policy preferences by viewing public and private documents and observing the foreign policy preferences of policy makers towards the perceived state. The policy preferences of the policy makers may or may not reflect the image perceived through the image indicators. Policy preferences reflect an action of some kind or, in some instances, a choice not to take action. Policy preferences for the dependent of the enemy would include diplomatic exchanges, covert actions, economic sanctions, and calls to the “enemy” to intercede in the crisis. Policy preferences towards the rogue state can be divided into retributive and rehabilitative policy actions. Retributive actions seek to uncover policies that reflect a prototypical rogue state image. Retributive actions would include economic sanctions, military actions or the use of force, blockades and coalition building. Rehabilitative policy preferences may reflect a less strict rogue state image, the desire for policy makers to promote regime change if instability is observed, or policy makers that do not perceive the state as a rogue at all. Rehabilitative actions would include diplomatic measures, the reduction of sanctions, and economic assistance. Policy preferences that could be either retributive or rehabilitative, or somewhere in the middle, include creating instability within the government structure, restrictive treaties, and containment strategies.

Policy preferences are the final decisions policy makers make towards the perceived state. In this way, policy preferences are collected after the origination or near the end of a crisis situation. Thus, there is a line between the collection of data for image indicators and the collection of data for policy preferences.

For the dependent variable, changes in the policy preferences of the policy maker are expected from one image to the other. Also, similarities and differences of policy makers’ policy preferences between Iraq and North Korea should occur. Part of this research is to determine what actions would be taken against a rogue state. I expect that as the rogue state increases in salience that the policy preferences of the perceiver state progresses towards expected response alternatives, including a greater use of force, restrictive treaties and the use of compellent strategies. In addition, I expect to observe a difference in degree of threat observed by the perceiver state and an obvious division between retributive and rehabilitative policy preferences.

I viewed three different points in time for Iraq and North Korea. Specific points of time were selected based on crisis situations where images could be formed. For Iraq, the first sample was taken during the years 1969 to 1973 for image indicators and 1973 to 1974 for policy preferences. During this time, a crisis situation occurred due to the Arab-Israeli war and the ensuing oil embargo that threatened the national interests of the United States. As oil is imperative to the mobilization of the military and the economic strength of the U.S., the crisis situation of the Arab-Israeli War and the subsequent oil embargo is ideal for observing the relationship between Iraq and the Soviet Union as perceived by U.S. policy makers. The second sample for Iraq was prior to the Iraqi invasion of Kuwait, specifically, the years 1989 to 1990 for image indicators, and then 1990 to 1991 for policy preferences. For the last sample, the period 1997 to 1998 prior to and during the Desert Fox campaign in December 1998, was chosen.

For North Korea, the first image indicator data was collected prior to the US Pueblo crisis in 1968. This presents the least supportive section of the study as there was a limited number of public remarks regarding North Korea prior to the Pueblo incident from which to obtain policy makers’ images of the state. Yet, there is enough data to get a good understanding of the relationship between North Korea and the Soviet Union. The policy preferences of policy makers develop through an analysis of the crisis beginning in January 1968 and ending in December 1968. The second sample of image indicators was collected prior to the crisis in 1993-1994 when North Korea withdrew from the Non-Proliferation Treaty and the period that followed. The difficulty for this time period was that the Clinton administration entered office January 1993, allowing for only a few short months to gather data before the onset of the crisis. I do include congressional statements prior to that time period, however, as there is more of a consistency with their holding office throughout this period of time. Policy preferences are gathered throughout the crisis starting in March, 1993 and ending in November, 1994. The last period for sample collection in North Korea was from 1997 to 2000, before and during the crisis instigated by the testing of the Taepo Dong missile by North Korea and a questioned underground facility within that state. The image indicators were collected prior to the onset of the crisis, while policy preferences developed from a sample collection once the crisis began.

As I hoped to observe the transition from the dependent of the enemy to the rogue state, a point before, during, and after the end of the Soviet Union was selected. The crisis point before the end of the Soviet Union should show a dependent of the enemy image. A crisis period around 1990 should show the beginning of the rogue state image, but there may be a transitional phase where a mixed type of image appears. Finally, a crisis point in the latter part of the 1990s should show a strong image of the rogue state. I expect that, as the rogue state image becomes more salient to the policy maker, policy preferences will reflect the changing image. Furthermore, I expect that, as the rogue state image becomes more salient and complex, policy makers’ policy preferences will reflect similarities and differences in the perceptions of Iraq and North Korea through retributive and rehabilitative policy preferences.

In both the Iraq and North Korea sample collections, an effort was made to separate the collection of data on image indicators and that of policy preferences. In some instances, this separation resulted in a less than desired number of coding units. However, I felt it better to draw a specific line between the data used for collecting the image indicators and the policy preferences than to blur the line and gain more image indicator coding units. In this way, every attempt was made not to create tautological data that simply reflected in the policy preferences the data obtained for image indicators. In addition, for the most part, the data on image indicators was collected, coded and analyzed prior to the collection of data for policy preferences, making the analysis of the image indicators a “clean” analysis without the prior knowledge of what policy preferences policy makers actually voiced during the crisis situation. The only instance where this did not occur was with the Pueblo incident, but I had collected all the data at one time nearly a year prior to the actual analysis of the data, and had only a vague remembrance of what policy makers actually voiced in their calls to action.

Policy makers share a conception of images used to structure the international system. In order to gain a clear picture of the arguments involved, I include a variety of policy makers. First is the executive branch: the President of the United States, the Secretary of State, the Secretary of Defense, the National Security Advisor and the CIA director. In all cases, the executive branch, as per the president, is held constant from image indicators to policy preferences. Thus, data is collected on the executive branch only those years (or months) prior to the crisis when the same administration is in office. In addition, I include national legislature members. I searched internet libraries (i.e. Lexus Nexus), official web sites (i.e. White House), the Congressional Record, the Department of State Bulletin, Vital Speeches of the Day and Public Papers of the Presidents. In addition, autobiographical texts were searched for statements that reflected images and policy preferences. While the data collection for the dependent of the enemy consisted of in depth library research with microfiche, micro film, primary documents and secondary documents, rogue state image data collection employed extensive internet research tools in additional to more traditional means.

I performed content analysis on primary documents, including interviews, speeches, public documents and press conferences, as well as secondary documents like autobiographies, in order to obtain the images held by policy makers and their policy preferences. Public statements provide information on policy makers for evaluation and analysis when data would otherwise be unavailable (Starr, 1984). As national leaders tend to be inaccessible to researchers and not likely to sit for a series of personality tests or fill out questionnaires, use of their statements in a systematic way provides the best solution for collecting useable data (Hermann, 1977). Statements by high level officials are often used to collect important data, especially when viewing personality, belief systems, and image creation (see Hermann, 1984; Holsti, 1977; Rosati, 1987). Ideally, the more spontaneous the statements, the more likely they will reflect the beliefs of the policy maker and not the beliefs of possible “ghost writers” (Hermann, 1984). Press interviews are an excellent source of spontaneous statements, as policy makers have less time to construct answers and will reflect their own beliefs (Hermann, 1984; Hermann, 1980b). In addition, House debates or the use of the Congressional Record, which tend to be more spontaneous statements or at least statements written by legislators, provide useful data for analysis (Hermann, 1977).

Yet, speeches, official statements and autobiographies should not be negated altogether. While some content analysis looks only at specific words, image theory looks more generally at the image portrayed, an image held by a group of individuals. There tends to be a “collective consciousness” within a group (Finaly, Holsti, and Fagen, 1967). This consciousness may be the same whether expressed by the policy maker or his/her speechwriter. In this way, a speechwriter would be presenting the image desired by the leader, an official statement reflecting the beliefs of the policy maker or the autobiographical sentiments of its writer. There tends to be a “prevailing view” of images used for foreign policy making, making speeches and statements of policy makers reported in the media acceptable sources for data collection (Herrmann, 1985a). In addition, while press statements may be more spontaneous than speeches and are less likely to reflect the writer’s beliefs over those of the leader, many leaders are prepped by their staff on what to say, an influence that could also be considered the views of staff rather than those of the policy maker. Furthermore, interviews and press conferences may be ideal sources of communication for data collection, but also tend to be limited in crisis situations (Holsti, 1976). Thus, speeches and autobiographies, when there is no other communications available, may still be useful in providing data on the images held by policy makers.

A caveat to any material communicated in writing or through statements is an acknowledgement that a particular audience exists to which the communication is directed, typically an appeal to a public audience (Holsti, 1976). Thus, any communication has the potential for being a political statement rather than a view of the individual’s beliefs. Yet, even the politicalization of an event reflects aspects of the policy maker’s personality and beliefs, observable through the images created and the policy preferences enacted.

As a method of research, content analysis has unique characteristics that allow answers to questions not perceivable through other means. Content analysis provides useful information on decision makers who are otherwise inaccessible to researchers (Hermann, 1977; Starr, 1984). While some questions and research designs allow for the use of questionnaires and personal interviews, other questions necessitate looking outside the light and searching a little in the dark to find possible answers. When data are limited or the actor is inaccessible (or no longer alive), content analysis provides the means through which research can still be done (Holsti, 1969). In international relations, answers often come from past events or are singular occasions that are rarely reproduced, often with few or no accessible participants. Content analysis overcomes this barrier and allows for the study of questions with answers lying in inaccessible events. Content analysis allows a researcher to take words communicated by policy makers and translate them into informative data, a form of inference (Starr, 1984; Holsti, 1969). Thus, inaccessible events are accessible through the use of content analysis. Furthermore, surveys tend to be intrusive and memories of individuals unreliable (Holsti, 1969). Content analysis allows data to be collected on statements when the event occurred or when researchers don’t want to invade the privacy of policy makers.

While the quantitative use of content analysis counts the frequency of words and is considered more systematic, generalizable, and theoretically relevant, qualitative content analysis provides the opportunity for researchers to infer a relationship between outcomes and events (Holsti, 1969). Political science and psychology both tend to use content analysis (Holsti, 1969), providing a rich history and support for its use in political psychology research. Both quantitative and qualitative content analysis should be used in research as they are part of a “continuum,” and qualitative content analysis may be best if it answers the “right” question (Holsti, 1969).

As with any good research method, content analysis is made better when certain rules are followed. Holsti (1969) defines content analysis as “any technique for making inferences by objectively and systematically identifying specified characteristics of messages” (14). Communications produce answers to questions through a deductive process whereby data are gathered and general concepts derived by reasoning. Such a process provides useful data when three criteria are followed: 1.) be objective, 2.) be systematic, and 3.) make sure the research is theoretically based (Holsti, 1969). Objectivity is relative. I have biases based on my gender, age, interest in international affairs and my unreasonable idealism that desires an end to all war. Yet, most of this can be put aside during the data collection and analysis process through the use of a systematic instrument based on image theory. In this way, all three criteria can be met, allowing for good research that can be replicated, built upon, and theoretically relevant.

The coding unit of analysis differs from the sample unit. Within each sample unit (each time period for collecting either the image indicators or the policy preferences), individuals communicate images and policy preferences that are collected and analyzed. These individuals are the coding unit of analysis. For Iraq in the sample units 1969 to 1973 (image indictors), 1973 to 1974 (policy preferences), 1989 to 1990 (image indicators), and 1990 to 1991 (policy preferences), and North Korea 1966 to 1968 (image indicators) and 1968 (policy preferences), all the statements of each individual over the sample unit of time were collected, colligated, and coded as a unit with the measure instrument. Each individual was a separate coding unit, providing an image for that individual. For Iraq 1996 to 1998 (image indicators), and 1998 (policy preferences), and North Korea 1992 to 1993 (image indicators), 1993 to 1994 (policy preferences), 1996 to 1998 (image indicators) and 1998 to 2000 (policy preferences), a computer software program was used to code the data per each speaker and speech. Instead of each speaker being the coding unit, the excerpt was the coding unit. Generally, those within the executive were grouped together as were those within the Legislature, although discrepancies between Republicans and Democrats were noted. Through this process, the n is increased, as per King, Keohane, and Verba.

The first measurement instrument (A) provides the structure through which the perceived image of states by individuals is obtained. Measurement instrument A concentrates on image indicators outlined in the independent variable. Measurement instrument A is used to determine the image held by the policy maker. Measurement instrument A can be found in the Appendix. The measurement instrument states the name of the policy maker, the individual’s title, and his/her party affiliation. Next, for those units coded by hand, the measurement instrument section (capability, culture and intent; event scripts; and response alternatives) observed in the documents is highlighted both on the measure instrument and in the documents. Capability is highlighted in yellow, culture in orange, intention in pink, events scripts in blue, and response alternatives in purple. In the margins, I made notes as to which indicators are present and to the degree in which the statements reflect the dependent of the enemy or rogue state imagery. If an image other than dependent of the enemy or rogue state was observed (i.e.: enemy image), this is also noted. Once the software program was available, the tedious work of notation ended. Content analysis still progressed through the reading of the excerpt (speech, Congressional Hearing, interview, etc.), but coding entailed the highlighting of the section to be coded by the mouse, and then clicking on the variable (or node) on a node tree (measurement instrument A) in the program (can be coded to multiple variables). A report was then produced for each variable (after the coding was completed for all excerpts) that showed each segment coded to that variable, who said it and when (and what ever other data I deemed necessary to that person), allowing for a more timely analysis of the data. Free nodes were also made that allowed for any unusual data to also be collected. None of the unique beauty of qualitative analysis was lost, but consistency and timeliness were improved. Indeed, coding became quite enjoyable.

The second measurement instrument (B) structures the policy preferences of the policy maker. Again, printed communications of each individual were collected and analyzed as a single coding unit. The second measurement instrument focuses on the policy preferences contained in the dependent variable. Measurement instrument B can be found in the Appendix. The measurement instrument states the name of the policy maker, the individual’s title, and his or her party affiliation. For those coded by hand, the policy preferences were highlighted, reflecting either policies for states perceived as dependents of the enemy or rogue states, and further divided into retributive or rehabilitative policy preferences, both on the measure instrument and in the documentation where observed. Retributive policy preferences were highlighted in green and rehabilitative policy preferences in blue. Policy preferences that could be either retributive or rehabilitative were highlighted in pink. Notes were also made in the margins next to the highlighted text logging the degree of response taken by the policy maker. For those units coded by the computer software, the highlight and click method of coding to a tree node was performed.

Through the use of the two measurement instruments, a systematic collection of data allows for replicability and theory building. The computer software program enhances the systematic process for coding data, significantly reduces the time needed for coding, but does not take away from the unique aspect of qualitative content analysis. In future research with a larger study, it can also help with coder reliability between coders.

[1] By beauty, I consider the greater depth of research on each case, the varied means for obtaining data, the descriptive aspect of some results, etc.

[2] As will be explained below, policy preferences are those actions that policy makers promote in a crisis situation.

Dependent of the Enemy Image	Rogue Image
inferior militarily	inferior militarily
inferior culturally	inferior culturally
inferior economically	inferior economically
support of "enemy's" military in creating deterrence	no deterrence from “enemy’s” military
	elite leadership
	anti-western policies
	aggressive
	seeks weapons of mass destruction
response alternatives:	response alternatives:
diplomatic exchanges	compellence
economic sanctions	use of force
covert actions	economic sanctions
calls to the “enemy” to intercede	create instability
	restrictive treaties