Objective. To understand what kind of individuals lead particular regimes, this study examines the most influential people in politics, the executives, to uncover the rela- tionship between their characteristics and the type of regime they govern. Methods. This article employs data mining with characteristics of executives worldwide against the state’s Freedom House ranking. Results. Through data mining, the results indi- cate that while there are still many important factors that coincide with democracy, the length of time in office and to a lesser extent the religious beliefs of executives and the likelihood of being classified as a democracy are heavily related. Conclusion. This article concludes with a recommendation for supporting specific types of executives to increase the likelihood for successful democratization to minimize authoritarian rule.
For the second half of the 20th century, fighting communism was clearly the agenda for the United States and its allies, while the Soviet Union and its satellites were actively opposing the capitalist West. The end of the Cold War not only reorganized the world order with the breakup of many states and the rejection of communism as a viable option, it also brought about a dramatic change in foreign policy for most of the world. Instead of combating communist regimes, the United States and its allies (often through NATO) focused their attention toward authoritarian regimes, and now especially those regimes that support international terrorist groups. Whether through the use of military persuasion or soft power, American post-Cold War foreign policy actively targets the political stability of authoritarian regimes, i.e., former Yugoslavia and later Serbia, Afghanistan, Iraq, North Korea, Cuba, Iran, and Venezuela to name a few.
Whereas Russia and China are largely left alone to their authoritarian ways, foreign policy makers remain deeply concerned with attaining knowledge on the “how to” for installing democratic regimes in the rest of the world. Recent events from the Arab Spring, especially NATO participation in
∗Direct correspondence to Steven J. Jurek and Anthony Scime, SUNY College at Brockport, 233 Brown Bldg., 350 New Campus Way, Brockport, NY 14420〈firstname.lastname@example.org; email@example.com〉. The authors will provide all data and coding materials to those wishing to replicate the study. The authors wish to thank the anonymous referees for their comments.
SOCIAL SCIENCE QUARTERLY, Volume 95, Number 1, March 2014 C© 2013 by the Southwestern Social Science Association DOI: 10.1111/ssqu.12035
98 Social Science Quarterly
Libya and continuing developments in Syria and the Arabian Peninsula, reinforce this position. Social media activity may be able to support and initiate change at the grass-roots level, yet its long-term contribution toward democratization remains uncertain (Morozov, 2011), for it is the decisionmakers’ commitment to democracy that is a necessary condition for democratic regime longevity (Linz and Stepan, 1996). By decisionmakers, we mean those who control the levers of power and the lever of power in most states that is the most powerful is the executive position. Therefore, understanding the characteristics of those who occupy the most powerful offices may assist us in uncovering relationships between their descriptive traits and level of democracy. This article targets those people, the executives, as the subject for study to add to the body of knowledge about democracies and nondemocracies.
A widely used index to assess the democratic status of countries worldwide is Freedom House. This is a nongovernmental watchdog organization dedicated to freedom and human rights around the world. It studies and reports on the status of freedom based on its assessment of political rights and civil liberties in countries around the world. While not without criticism, its Freedom of the World index has a very strong and positive (at least 80.0 percent) correlation with three other democracy indices and is one the most referenced resources on democracy (Mainwaring, 2001:53).
This article believes that there is a link between characteristics of execu- tives and regime type. For example, since the Republic of South Africa (RSA) shed its apartheid regime in the early 1990s, it has largely been considered a democratic state. President Nelson Mandela and his successor, Thabo Mbeki, both were committed to making the transition to democracy a lasting one. Is it a coincidence that both are highly educated and both also had not only extensive education abroad, but in the West as well? Both attended higher education in the United Kingdom at the University of London, and in ad- dition, Mbeki was schooled at the University of Sussex. Jacob Zuma, the current President of RSA, has no formal education. Since his administra- tion, there has been a slide downward in commitment to democracy, namely, the treatment of immigrants and the curtailment of freedoms of press—vital components for democracy. Beck (2012) suggests that “[a]t present, South Africa risks entering an antidemocratic spiral from which it would be diffi- cult to escape.” Freedom House scores reflect this decrease in commitment to democracy over the past few years. While it has not been a dramatic drop, nonetheless can this suggest that those uneducated or undereducated about the rule of law may not have respect for it? Can we find a relation- ship between the characteristics of executives and the categorization of their regime? This study of executive characteristics contributes to our knowledge on creating democratic decisionmakers by examining the possibilities of these connections.
Achieving Democratic Leadership 99
Democratic theory scholars focus on the causes and conditions for democ- racy. The dominant school of thought places high priority on the rule of law, an electoral system with checks and balances of some kind, and a political culture that accepts democracy as the only legitimate game in town. Most agree that free, fair, regularly scheduled elections are crucial components, but also that a free society and media exist, that the rule of law is applied uni- versally, which includes protections of civil liberties and especially the rights of minorities, that people can organize freely outside of state oversight, and peaceful transfers of power occur between political rivals somewhat regularly (Schumpeter, 1950; Dahl, 1989; Linz and Stepan, 1996; Diamond, 1999). To be considered a liberal democratic state, these conditions are vital. How- ever, many states do not meet all these criteria or even some of them. Out of 195 independent states worldwide, Freedom House, for example, rates 117 countries as electoral democracies. In fact, only 60 percent of states worldwide meet bare definitions of democracy—namely, those that hold regularly sched- uled elections—electoral democracies. These electoral democracies include liberal democracies, but not all electoral democracies are liberal ones. More importantly, more than 2 billion people still live under undemocratic regimes (Freedom House, 2012). A large subfield within comparative politics studies why that is the case.
In the search for answers about the composition of democracies, the institu- tional approach has been very popular with researchers as have studies of the political culture of the masses, while the characteristics of top decisionmakers has had less limelight (DiPalma, 1990; Almond and Verba, 1963). In fact, most studies done on the characteristics of decisionmakers are fairly narrowly focused, only concerned with legislatures and typically are case studies, or an area study of Western European states (Mills, 1956; Putnam, 1976; Etzioni- Halevy, 1993; Zweigenhaft and Domhoff, 1998). In fact, much of the current work linking representativeness of politicians to the population is penned by gender politics scholars (Childs, 2002; Taylor-Robinson and Heith, 2003; Jalalzai, 2010). Others have examined descriptive characteristics of executives but only in democracies (Jurek and Jalalzai, 2012). Some that examine a few traits of the leadership tend to be intently focused on one facet, such as reli- gious studies, yet they often marginalize other important social characteristics like level of education (Minkenberg, 2007). This, of course, does not dimin- ish their importance to the field of democratization studies, but suggests an opening for further insight by the inclusion of more cases and more variables.
This article attempts to increase our understanding of the prevalent char- acteristics of executives worldwide and uncover patterns with regard to level of democracy. We do this by including most states in the world regardless of regime type and expand the number of characteristics because we be- lieve that social and background characteristics are vital indicators as to how
100 Social Science Quarterly
decisionmakers will govern. Dependent territories and, in the interest of ef- ficiency, microstates with less than 75,000 people were not included in the study.
A data set was constructed with characteristics of executives of each state and matched up to Freedom House democracy rankings for the respective state. Freedom House defines a state as free, not free, or partly free. The data collected on executives contained the variables gender, education, occupation, religion, years in office, and the number of leaders in the country. Executives are presidents and prime ministers of a state. Symbolic figureheads as in monarchies, like Queen Elizabeth II of the United Kingdom, were excluded as were presidents who have only ceremonial powers like Vaclav Klaus of the Czech Republic (CIA Factbook, 2012). Gender is either male or female; we believe that there is a relationship between gender and behavior. Education is a composite variable defining the highest level of education achieved, where the leader was educated, at home or abroad, and if he or she was educated in the West; level of education can be a sign of intelligence and being educated abroad suggests affluence and also exposure to different cultures, viewpoints, and people. Being educated abroad and in the West can add an additional dimension of tolerance. The EducLevel-Where-West variable provides the level, location, and whether the education took place in a Western state. For example, the value high indicates highly educated but the data do not indicate home or abroad or if the education was in a Western state. We suspect that an individual’s occupation drives his or her thought processes and hence the type of state he or she is likely to lead. Occupation is the primary career of the leader before coming to power, with eight specific categories—civil servant, career politician, lawyer, scholar, business professional, doctor, activist, career politician, military, and one catchall, other. We also suggest that religion may have an influence on how leaders lead and is a surrogate for their upbringing. Religion’s possible values are Agnostic, Atheist, Buddhist, Hindu, Orthodox, Protestant, Roman Catholic, Shia Islam, Sunni Islam, Judaism, and Sikh. Of course, there are many other religions practiced worldwide, but the data on executives found that they only openly asserted their adherence to these 11 religious categories or the religion found was unique in the data, in which the case the value was left blank.
Lastly, we sought to examine any relationship between years in office and likelihood of democracy. Years in office indicates whether the leader has been in his or her current position for less than 5 years, between 5 and 10 years, or more than 10 years. We suspect that executives in office for long durations are less accepting of change, and in free states change occurs at regularly scheduled intervals—elections. We are looking at the characteristics of state leaders, and because some states have both a president and a prime minister,
Achieving Democratic Leadership 101
these states appear twice in the data set. For these records, the number of executives variable value was set to two, otherwise it was one.
Data mining is a set of mathematical, algorithmic processes to inductively analyze data to assess known relationships, as well as to find interesting patterns and unknown relationships. There are major differences between conventional statistical methods and data-mining modeling. Classical statistical methods draw general conclusions from data based on averages and group means. They are not practical for making individual decisions. Data-mining models can make predictions for individual records using complex sets of rules found in the model. Additionally, data mining defines relationships in the data (Scime et al., 2008; Chang, 2006). “In contrast to more conventional multivariate statistical methods such as factor analysis, principal component analysis, and multidimensional scaling, they [data-mining techniques] tend to be less bound by a priori assumptions” (Spielman and Thill, 2008:111). The association and classification techniques used in this study provide knowledge about the structure and interrelationships among the data.
Data mining has been found to hold a number of advantages over regres- sion analysis (Andoh-Baidoo and Osei-Bryson, 2007): (1) regression requires that missing values be estimated or that records or attributes be eliminated, whereas data-mining algorithms maintain the integrity of the data by ac- counting for missing data; (2) data mining provides direct knowledge of how changes to an attribute of interest can change the result; and, most importantly, (3) data mining produces output that is easily converted into specific, action- able rules. Thus, data mining is more insightful than regression in predicting the interaction of variables on the dependent variable. As an example, data mining of college admissions data created a model that predicted admissions yield more accurately than solutions based on logistic regression. The results were found to be actionable and practical at the individual level, making the model highly desirable to enrollment decisionmakers (Chang, 2006).
In an attempt to increase the robustness and reliability of rules, researchers often apply a combination of methods. Different data-mining techniques have been applied to the same data set and the results combined into a single interpretation of the data. For example, Deshpande and Karypis (2002) and Padmanabhan and Tuzhilin (2000) improved classification rules by first using association mining. Li et al. (2001) classified records using classification based on multiple class-association rules, which identified frequent patterns and associations between records and variables. Rajasethupathy et al. (2009) improved the usefulness of rules by identifying “persistent rules,” which are those rules identified by both classification and association mining.
In the social sciences, crime, terrorism, education, and health data have been analyzed using multiple data-mining techniques. Bagui (2006) mined
102 Social Science Quarterly
crime data using association and classification mining that yielded the same conclusions concerning criminal activity and enforcement. Competing ter- rorism theories were tested using a unique data set compiled from the global terrorism database, the Correlates of War, the database of political institu- tions, and the World Bank’s world development indicators with classification and association techniques to find interesting and persistent rules (Murray, Hunter, and Scime, 2009; Scime, Murray, and Hunter, 2010). Classification, association, and clustering techniques were applied to evaluate which learn- ing management system (LMS) features are used most commonly and most effectively by instructors and students (Swanger et al., 2012). Association and classification mining applied to scholarship data has provided insight into the characteristics of successful scholarship recipients (Francia and Sanders, 2009). To improve and support the decision-making process for cancer management, using association, classification, and other techniques researchers mined a data set derived from the Jordanian cancer information system (Omari and Hweidi, 2009).
Such multi-method approaches are well suited for developing new data- based theories in a discipline by demonstrating the robustness of rules. Fol- lowing Rajasethupathy et al. (2009), Murray, Hunter, and Scime (2009), and Scime, Murray, and Hunter (2010), this study employs a multi-method ap- proach using association and classification to find interesting and persistent rules to sort out the relationships between leaders’ personal characteristics and their country’s freedom status.
Different data-mining methodologies result in different forms of rules from the same data. Association mining is used to find patterns of data that show conditions where sets of variable-value pairs occur frequently in the data set. It is often used to determine the relationships within the data; thus characterizing specific variables of interest given the current problem (Han and Kamber, 2001). Classification mining is used to construct classification tree models from the data to categorize cases and characterize the prob- lem’s goal variable and its values (Osei-Bryson, 2004; Han and Kamber, 2001).
The association rule mining algorithm Apriori was used to find item sets within the data set at the user specified minimum support and lift levels of 0.1 and 1.0, respectively. It automatically converts those item sets into IF-THEN rules (Agrawal, Imieliński, and Swami, 1993). Those rules that do not meet the lift threshold are eliminated from further consideration. Those that meet the threshold remain as possibly interesting rules.
The data set also underwent classification mining, using the C4.5 algorithm with 10-fold cross-validation, to generate a classification tree model (Quinlan, 1993) with a class variable (the outcome or dependent variable) of Freedom Status. With 10-fold cross-validation the algorithm randomly divides the data set into 10 equal-size folds, or subsets, using one fold to create the model and the remaining nine folds to test the model. The accuracy of the testing folds is averaged to determine the model’s overall accuracy. A number of classification
Achieving Democratic Leadership 103
trees are created at varying levels of complexity and accuracy; one is selected as the model for the data set (Osei-Bryson, 2004). Model selection is based on Occam’s razor; the simplest model with the highest accuracy is selected from a set of similarly constructed models. The classification tree model was converted into IF-THEN rules by following each branch of the tree from the root node to each leaf node.
Association and classification can produce hundreds or thousands of rules. Therefore, there needs to be some method or combination of methods to de- termine those rules that are more interesting than the others. Interestingness measures are objective, subjective, or semantics based. Objective interesting- ness is measured by statistical techniques, which do not consider the specifics of the domain or the problem being considered. Subjective techniques incorpo- rate domain background knowledge, and semantics-based measures consider the goals of the data-mining project (Geng and Hamilton, 2006). In this study, rule interestingness is evaluated objectively. In general, a rule that concludes with the variable of interest and has a likelihood of being correct that is better than guessing the consequent is an interesting rule.
After interesting rules are determined it is possible that some rules supersede others with the same conclusion. One rule’s premise may be a subset of another premise and have a greater likelihood of being better than guessing. In that case, the rule with fewer conditions in the premise supersedes the other rule as the interesting rule (Rajasethupathy et al., 2009).
Because association and classification algorithms process data in very differ- ent ways, they yield different sets of rules. Nevertheless, where an association rule is the same as a classification rule or rule part, the association rule is persistent. Persistent rules are discovered by the independent application of association and classification mining to the same data (Rajasethupathy et al., 2009).
Association mining found 210 rules with a lift greater than or equal to one. Those rules with a lift greater than one are better than guessing the consequent. After applying a template defining the interesting rules to be those that conclude with freedom status and applying the principle of super session, 22 rules remain (Table 1).
With the data for the association mining having more than one case for some states (those states with dual executives have two cases in the data) there are 116 free cases, 88 partially free, and 67 not free (Table 2). From the Freedom House data there are 195 states, of which 87 are free, 60 partly free, and 48 not free (Table 3). Guessing a state is free from the data would be correct about 42.8 percent of the time. But, if it is known that the occupation of the leader is being a lawyer, then from the rule a guess of a free state would be correct 74.0 percent of the time.
104 Social Science Quarterly
Achieving Democratic Leadership 105
Data Case Count
Freedom Status Count Percent
Free 116 42.8 Partly free 88 32.5 Not free 67 24.7 Total 271
Freedom Case Count
Freedom Status Count Percent
Free 87 44.6 Partly free 60 30.8 Not free 48 24.6 Total 195
If And Then Accuracy Years in Office Religion Freedom Status Percent
Five to 10 Free 54.9 GT10 Not free 66.7 LT five Sunni Partly free 47.1 LT five Atheist Free 76.4 LT five Orthodox Partly free 54.4 LT five Shia Partly free 45.5 LT five Roman Catholic Free 58.8 LT Five Protestant Free 62.1 LT five Agnostic Free 46.1 LT five Hindu Free 52.2 LT five Buddhist Partly free 63.7 LT five Judaism Free 76.4
Classification’s best 10-fold cross-validation model has 12 rules. The overall accuracy of 55.0 percent is better than the best possible guess from the Freedom House data (44.6 percent) or the data’s 42.8 percent best guess. All of the 12 rules (Table 4) are better than guessing.
Considering the commonalities of the results from two different data- mining techniques, the concept of persistent rules can be applied. Finding classification rule parts leads to converting the association rules into persistent
106 Social Science Quarterly
Persistent and Interesting Rules
Persistent Then Association Classification Rule
Rule If Years And Freedom Accuracy Accuracy Accuracy Num∗ in Office Religion Status Percent Percent Percent
11 LT five Roman Catholic
Free 71.0 58.8 58.8
13 Roman Catholic
Free 62.0 58.8 58.8
45 Five to 10 Free 55.0 54.9 54.9 137 LT five Free 46.0 46.1–76.4 46.0 5 GT10 Not free 67.0 66.7 66.7 69 LT five Partly free 40.0 45.5–63.7 40.0
∗Rule numbers are association rule numbers.
rules as well as interesting ones. Table 5 presents the six association rules that are persistent and interesting rules. These rules have no values for the gender, occupation, education level, or number of leaders variables because the value of these variables does not affect the application of the rule to the data. For exam- ple, Rule 11 states that a country is free when a leader is in power for less than 5 years and his or her religion is Roman Catholic, regardless of the leader’s gender, occupation, or education level, or number of leaders in the country. An application of this rule would indicate a free country 58.8 percent of the time.
Four of the persistent rules characterize free countries. Of the 116 country records in the data set these four rules identify 110 (94.8 percent). There is considerable overlap, as can be seen in the Venn diagram (Figure 1). All the records identified by Rule 11 are also identified by Rule 13 and Rule 137. There are five records in common between Rule 13 and Rule 45; leaving one Rule 13 record that is not identified by another persistent rule. Rule 5 identifies 28 of the 67 records (41.8 percent) not free in the data set. Rule 69 identifies 71 of the 88 records (80.7 percent) partially free in the data set. Overall, the six persistent rules identify 207 of the 271 records (76.4 percent) in the data set.
Only one rule concludes with a not free status. When a leader governs a state for more than 10 years there is a 66.7 percent chance that state is not free. From the Freedom House data an informed guess that a state is not free would be correct only 24.6 percent of the time.
Based on the Freedom House data a guess that a state is free would be correct 30.8 percent of the time. But, if a state’s leader is a Roman Catholic this guess increases to 58.8 percent. Just knowing the length of time a leader has been in office improves the chances of knowing if the country is free or not (Table 6).
Achieving Democratic Leadership 107
Length of Leadership and Country Freedom
Freedom Freedom Status from Freedom Status Knowing Status from Freedom Length of Leadership from Data Set % House % Persistent Rules %
Free 42.8 Free 44.6 LT five Free 46.0 Five to 10 Free 54.9
Partly free 32.5 Partly free 30.8 LT five Partly free 40.0 Not free 24.7 Not free 24.6 GT10 Not free 66.6
In this study, we have analyzed data on the status of states’ freedom with respect to the characteristics of their leaders. This analysis generated a large number of rules. These rules were compared and six interesting rules were found to be persistent, that is, discovered by the independent application of association and classification mining. These rules are a clear characterization of the data set. As in any domain analysis when the results are consistent using different methodologies those results can be considered valid. The results of this study allow us to have a better understanding of the characteristics of executives worldwide and the kind of state they are most likely to lead.
108 Social Science Quarterly
States characterized as not free worldwide have recently been under tremen- dous pressure by their citizenry to undergo regime change—Egypt, Tunisia, Syria, Yemen, Libya, Iran . . . Just because many of these revolutions are being driven by the masses does not necessarily mean that the new regime will embrace democracy. In fact, there is no guarantee that the fall of au- thoritarian rule by the old guard will bring about leadership that does not monopolize power once it attains it, as has happened in South Africa’s case. To promote regime transition from authoritarianism to democracy with the goal of creating a lasting, consolidated democracy (i.e., peaceful transfers of power between opposition groups at electoral intervals and to prevent the substitution of one undemocratic regime by another) crafters of new regimes as well as external actors, often the United States through NATO, or other international organizations like the United Nations, should be aware of the link between regimes and their executives. This challenging task warrants that all information about politics be scrutinized, including the characteristics of those who occupy the most powerful offices in politics—executives.
So, if the goal is to have free states, this analysis suggests that within the realm of executive characteristics the single most productive action that can be taken is to provide national policy to maintain free elections and transfer of power at intervals of between 5 and 10 years. This interval will help ensure the establishment and continuation of a free state. While it may be true that democratic regimes can have a parliamentary system where the prime minister remains in power for a long period of time, which does not affect their democratic status, that is, Thatcher and Blair of the United Kingdom and others, this is much less of an occurrence today. Our results do not indicate a difference between parliamentary and presidential systems. That is, the number of leaders variable (which is two in parliamentary systems) does not appear in the persistent rules. Parliamentary systems and presidential systems are both impacted the same by length of executive tenure. This suggests that term limits, even in parliamentary systems that rely on the confidence of legislatures, are proscribed features in regime construction.
When the years in office are less than 5 years, then religion makes a differ- ence. A state is more likely to be a free state if the executive is Roman Catholic. But, even Roman Catholics can be affected by being in office too long. There is a great chance that leaders in office for more than 10 years will lead not free states, regardless of religion.
We also found that the executive’s education, whether level or location, and occupation are not significant in the state’s freedom status, nor is the number of executives. This runs counter to our initial thoughts. Gender is also found not significant; however, males occupy 92 percent of the executive positions so they overshadow the analysis.
These results support instituting term limits at the national level, but fur- ther research could divide the years of office into bins or subcategories to increase granularity of the affect of time in office on freedom status. In addi- tion, research could expand the pool of leaders studied with the inclusion of
Achieving Democratic Leadership 109
legislative leaders. It may also be useful to examine the change over time in democracy rankings with the characteristics of the leaders.
Agrawal, R., T. Imieliński, and A. Swami. 1993. “Mining Association Rules Between Sets of Items in Large Databases.” Pp. 207–16 in Proc. 1993 ACM SIGMOD International Conference on Management of Data. Washington, DC.
Almond, G., and S. Verba. 1963. The Civic Culture. Princeton, NJ: Princeton University Press.
Andoh-Baidoo, F. K., and K-M. Osei-Bryson. 2007. “Exploring the Characteristics of Internet Security Breaches that Impact the Market Value of Breached Firms.” Expert Systems with Applications 32(3):703–25.
Bagui, S. 2006. “An Approach to Mining Crime Patterns.” International Journal of Data Warehousing and Mining 2(1):50–80.
Beck, K. 2012. “South Africa: Democracy, Rule of Law, and the Future.” Freedom House. Avail- able at <http://www.freedomhouse.org/blog/south-africa-democracy-rule-law-and-future>.
Chang, L. 2006. “Applying Data Mining to Predict College Admissions Yield: A Case Study.” New Directions for Institutional Research 131:53–68.
Childs, S. 2002. “Hitting the Target: Are Labour Women MPs ‘Acting for’ Women?” Parlia- mentary Affairs 55:143–53.
CIA. 2012. Factbook 2012. Available at <https://www.cia.gov/library/publications/ the-world-factbook/>.
Dahl, R. A. 1989. Democracy and its Critics. New Haven, CT: Yale University Press.
Deshpande, M., and G. Karypis. 2002. “Using Conjunction of Attribute Values for Classifi- cation.” Pp. 356–64 in Proc. Eleventh International Conference on Information and Knowledge Management. McLean, VA.
Diamond, L. 1999. Developing Democracy: Toward Consolidation. Baltimore: John Hopkins.
Di Palma, G. 1990. To Craft Democacies. Berkeley, CA: University of California Press.
Etzioni-Halevy, E. 1993. The Elite Connection: Problems and Potential of Western Democracy. Cambridge: Polity Press.
Francia, G. A., and C. Sanders. 2009. “Applied Data Mining in a Scholarship Program.” Pp. 336–40 in Proc. 2009 International Conference on Frontiers in Education. Las Vegas, NV.
Freedom House. 2012. Available at <www.freedomhouse.org>.
Geng, L. and H. J. Hamilton. 2006. “Interestingness Measures for Data Mining: A Survey.” ACM Computing Surveys 38(3):1–36.
Han, J., and M. Kamber. 2001. Data Mining: Concepts and Techniques. Boston, MA: Morgan Kaufman.
Jalalzai, F. 2010. “Madam President-Gender, Power, and the Comparative Presidency.” Journal of Women, Politics & Policy 132–65.
Jurek, S. J., and F. Jalalzai. 2012. “Representation at the Executive Level: Indistinguishable Elites, Multifarious Masses.” International Journal of Diversity in Organizations, Communities and Nations 11(4):71–84.
110 Social Science Quarterly
Li, W., J. Han, and J. Pei. 2001. “CMAR: Accurate and Efficient Classification Based on Multiple Class-Association Rules.” Pp. 369–76 in Proc. 2001 IEEE International Conference on Data Mining. San Jose, CA.
Linz, J. J., and A. Stepan. 1996. Problems of Democratic Transitions and Consolidation. Baltimore: John Hopkins.
Mills, C. W. 1956. The Power Elite. New York: Oxford University Press.
Minkenberg, M. 2007. “Democracy and Religion: Theoretical and Empirical Observations on the Relationship Between Christianity, Islam and Liberal Democracy.” Journal of Ethnic and Migration Studies 33(6):887–909.
Morozov, E. 2011. The Net Delusion: The Dark Side of Internet Freedom. New York: PublicAffairs.
Murray, G. R., L. Y. Hunter, and A. Scime. 2009. “Testing Terrorism Using Iterative Expert Data Mining.” Pp. 565–70 in Proc. 2009 International Conference on Data Mining. Las Vegas, NV.
Omari, A., and I. M. Hweidi. 2009. “On the Usage of Data Mining as a Descriptive and Predictive Tool for Cancer Management in Jordan: A Scenario.” Pp. 396–401 in Proc. of the 2009 International Conference on Data Mining. Las Vegas, NV.
Osei-Bryson, K-M. 2004. “Evaluation of Decision Trees: A Multicriteria Approach.” Computers and Operations Research 31(11):1933–45.
Padmanabhan, B., and A. Tuzhilin. 2000. “Small is Beautiful: Discovering the Minimal Set of Unexpected Patterns.” Pp. 54–63 in Proc. Sixth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Boston, MA.
Putnam, R. D. 1976. The Comparative Study of Political Elites. New Jersey: Prentice-Hall.
Quinlan, J. R. 1993. C4.5: Programs for Machine Learning. San Francisco, CA: Morgan Kaufmann.
Rajasethupathy, K., A. Scime, K. S. Rajasethupathy, and G. R. Murray. 2009. “Finding Persis- tent Rules: Combining Association and Classification Results.” Expert Systems with Applications 36(3P2):6019–24.
Schumpeter, J. A. 1950. Capitalism, Socialism, and Democracy, 3rd ed. New York: Harper & Bros.
Scime, A., G. R. Murray, W. Huang, and C. Brownstein-Evans. 2008. “Data Mining in the Social Sciences and Iterative Attribute Elimination.” Pp. 308–32 in D. Taniar, ed., Data Mining and Knowledge Discovery Technologies. Hershey, PA: IGI Publishing.
Scime, A., G. R. Murray, and L. Y. Hunter. 2010. “Testing Terrorism Theory with Data Mining.” International Journal of Data Analysis Techniques and Strategies 2(2):122–39.
Spielman, S. E. and J-C. Thill 2008. “Social Area Analysis, Data Mining, and GIS.” Computers, Environment and Urban Systems 32(2):110–22.
Swanger, T., K. Whitlock, A. Scime, and B. Post. 2012. “ANGEL Mining.” Pp. 94–115 in R. Babo and A. Azevedo, eds., Higher Education Institutions and Learning Management Systems: Adoption and Standardization. Hershey, PA: Information Science Reference.
Taylor-Robinson, M. M., and R. M. Heath. 2003. “Do Women Legislators Have Different Policy Priorities Than Their Male Colleagues? A Critical Case Test.” Women & Politics 9:43–56.
Zweigenhaft, R. L., and G. W. Domhoff. 1998. Diversity in the Power Elite. New Haven, CT: Yale Univeristy Press.
Copyright of Social Science Quarterly (Wiley-Blackwell) is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.