A Review of "Queens" - Does the Gender of Leaders Impact Conflict Participation

By: Julian Olsen-Pendergast

 ‘Queens' by Oeindrila Dube et al. aims to determine whether systematic differences exist between male and female leaders and, if so, whether they impact the institutions they govern and the policies they implement. This is determined through the examination of the historical rulers of Europe from the fifteenth to twentieth centuries and conflicts that occurred under their rule. Specifically through examining whether polities[1] led by a female leader, as opposed to a male leader, are more likely to participate in conflict, a decision in which the leader of the polity plays a crucial role. Differing on a conceptual level from the question of whether women are more violent than men, as leaders who employ pro-conflict policies are not only influenced by their disposition to violent acts. The paper builds upon two theories that explain why female leaders might be involved in more conflicts. Firstly, due to their perceived gender weakness, making surrounding polities view them as an 'easy target' (Beem, 2014). Secondly, the purposeful involvement in conflict as a means to unite their subjects under a common enemy, thereby reducing internal instability (Ostrom, 1986).

Few studies have examined the relationship between the gender of leaders and conflict, primarily because of the historical scarcity of female leaders. As a result, previous research faced challenges in estimating the impact of gender on involvement in conflicts. This review will examine the econometric methods used by the authors, as well as critically evaluating the papers strengths and weaknesses.

The paper establishes three central findings. Firstly, unwed queens are more likely to be involved in a conflict pursued against them than unwed kings. Secondly, wed queens are more likely to pursue a foreign campaign than wed kings. Lastly, the authors theorise that the reason wed queens are more likely to be involved in conflict is due to the division of labour theory. The theory states that historically, wed queens often appointed their spouse as a prominent figure in their military leadership. Given the data from the study was extracted from an era of patriarchy, it is thought that the lower-status husband would be more likely to pursue aggressive foreign military campaigns as a means to boost his status.

With no pre-existing dataset containing the required data for this analysis, the authors constructed a new dataset from various sources. The dataset begins when the data became readily available, at the end of the 15th century, and ends at the emergence of the First World War. They justified this cut-off due to the First World War symbolizing the end of most of Europe’s monarchs' power in deciding whether their polity should be involved in a conflict. The genealogy data of the monarchies between 1480–1913, collected from 'Dynasties of the World' and 'Catalog of Royal Family Lineages' (Morby, 1989; Tompsett 1994). The second category was the data on wars and conflicts, where the authors used 'A Study of War' (Wright, 1983). The last category of data collected was other measures of stability and territorial expansion. This data was collected from 'Killing Kings: Patterns of Regicide in Europe, AD 600–1800' and 'Dynasties of the World' (Eisner, 2011; Morby, 1989). This data collection left the authors with a main sample of 18 polities that had a queen in the period between 1480–1913. Additionally, an auxiliary sample of 18 polities which never had a ruling queen was added. Leaving the authors with 3,586 observations.

 Using an instrumental variables (IV) strategy, Dube et al. were able to estimate the effect of queens on polity conflict participation. At the first stage they used two instruments; First, the presence of a male firstborn child, and second, whether the previous monarch had a sister. Thereby determining the likelihood of a queen ascending to power. At the second stage the authors established the effect of a polity ruled by a queen on engagement in conflict during a particular year. Dube et al. applied additional specifications, including decade fixed effects and more.

The authors validated the use of an instrumental variable for the gender of the firstborn child for four reasons. Firstly, due to the arbitrariness of a child's gender, they plausibly assumed that the instrument would be exogenous to conflict participation. The secondary motivation was to eliminate potential biases related to the efforts of monarchs in determining the gender of their heirs. As actively seeking a male heir might be linked to aggressive behaviours. Thirdly, there was concern about illegitimate heirs, as the efforts to eliminate illegitimate heirs might impact conflict tendencies. However, the authors noted that their ascension to rule is statistically unprobeable, and the gender of the firstborn child, whether legitimate or illegitimate, was found not to promote conflict. Lastly, it was used to invalidate concerns about the potential misclassification of illegitimate sons as legitimate in the data. Through robustness checks they found that these concerns of data misclassification had no meaningful effect on the results.

 Next the authors address the issues surrounding the samples selection bias for potential heirs’ survival to age of ascension. They theorise that the children that survive to ascension may be more aggressive, as if they are to survive they must overcome harsh living conditions or aggressive competition among siblings, including participation in warfare. To correct this issue Dube et al. exploit the legitimate firstborn’s gender, as opposed to the oldest legitimate surviving child at ascension. Furthermore, the authors also employ an auxiliary control measures for deceased siblings.

The authors further validated the sample instruments precision through comparing the sample sex ratio at birth with a present day naturally occurring ratio. They found that their sample was consistent to the current naturally occurring ratio (within a margin of error). Disproving the idea that polity rulers participated in sex-selective infanticide. Thereby, addressing the last concerns of selection biases and emphasising the completeness and accuracy of the sample's genealogical data.

 Additionally, Dube et al. implement control flexibility for the number of siblings a previous monarch had, to account for potential confounding factors. Evidently, the previous monarch having a sister is contingent on the monarch of two periods prior having a daughter. This monarch from two periods ago would be more likely to have a daughter if they had a high number of children, posing challenges due to the higher number of children that could potentially bid for ascension. Thereby introducing a pathway for future conflict. Additionally, Dube et al. implemented controls for deceased children due to early deaths from conflict, and for cases where the firstborn's gender is missing from records. Through these implementations the authors justifiably enhanced the robustness of their analysis.

Lastly, when interpreting the instrumental variable estimates, the authors acknowledged the limitations of the approach, as the cases in which a queen ascending to rule is dependent on arbitrary factors outside the gender of their siblings. Furthermore, they found that there was potential for heterogeneity in treatment effects, as it will only have an impact on the specific cases where a woman was eligible to rule. However, the authors found that the results are similar across the instrument sets, reducing the significance of claims of potential for heterogeneity in treatment effects.

 The paper successfully presents their evidence on how queens affect conflict participation. Table 3 examines the Ordinary Least Squares (OLS) and IV relationships between queens and conflict participation. The authors found through OLS that if a queen was the ruling monarchy of a polity that that polity would be 13 percentage points more likely to be involved in a conflict than if they had a king. However, the OLS estimates might have problems concerning downward biases, as queens might be more likely to be allowed to rule during times of stability. To account for these potential biases they present the IV estimates. Finding that when using instruments for the previous monarch having a sister and presence of a first-born male, a queen ruled polity was 39 percentage points more likely to be involved in a conflict than a king ruled polity. Additionally, the IV estimates for a queen ascending to power if the previous monarchy had a first-born male as negative, and previous monarch having a sister was positive. Which is consistent with what we would expect in the sample.

The authors present an array of additional validity checks on the instrument set and robustness checks. The instrumental validity held as table 4 shows that even in the few cases in which attempts to crown an illegitimate male heir as opposed to the firstborn female heir did not cause further conflict. Moreover, the authors provided additional controls and sensitivity checks due to the limited number of queens presented in the sample. The authors found that the effects are not predominately driven by minor cases of infrequent queenly rule and remain in place under alternative specifications.

Dube et al. took their analysis a step further by exploring the accounts of why these effects on conflict arise. Their results on conflict type are presented in table 10 and suggest that queens did not enter conflicts purely when being attacked. Disproving the ‘easy target theory (Beam, 2014). The results from table 5 show that a queen led polities were 36 percentage points more likely to enter a new war than king led polity, with the continuation of previous wars having little effect on the overall queen participation in conflict. Additionally, they found that queens where 40 percentage points more likely to be the aggressors in the conflicts created.

To examine the heterogeneous effects of marital status that were previously mentioned in the authors theory of division of labour in royal marriages, they incorporated a Married in Reign variable with the Queen indicator. The results of which can be found in table 6, where the authors found that a married queen was 70 percentage points more likely to be the attacking polity in a conflict than a king. Additionally, finding that unwed queens where 42 percentage points more likely to be the attacked polity as opposed to king led polities. However, a potential endogeneity concern was raised in relation to the specification of marriages being organized for strategic reasons, and therefore selecting abnormally belligerent male suitors with expansionist agendas for marriage to the queen. Overall the results suggested the theory of asymmetries in the division of labour in queenly marriages to be valid.

Lastly, three alternative accounts to these findings where further explored by the authors; The first alternative account was queens attacking to show their strength to neighbouring polities. However, the authors found that this account holds no significance as if that was to be the case one would see queens attacking at the start of their reins, which the results in table 8 disprove. If anything they were more prone to attack towards the end of their rein. The second alternative account was queens pursuing external conflicts as a means to unify the polity against a common enemy, thereby reducing internal instability. However, the results from table 7 show that having a queen as the ruling monarch of a polity does not bring about its demise, thus proving this alternative account has no significant effects on a queen’s participation in conflict. The last alternative account discussed was that of queens’ aggressive conflict strategies being the product of recommendations of advisors and foreign ministers. This account only holds under the assumption that advisors have more aggressive foreign policies than queens, and that queens are more easily influenced by advisors than kings. However, the authors disproved the account. Finding that queens who ascended to the throne at a younger age were less likely to be involved in conflict. In a case where the account was to be true, one would expect conflict effects to be larger if they ascended to the throne at a younger age, but this was not the case.

 In conclusion, the paper is successful in its aim to examine the relationship between polities ruled by a queen and conflict participation. The authors found that polities ruled by a queen were more likely to be involved in conflict than kings, with those that are married more likely to be the attacking polity, validating their division on labour theory. They successfully acknowledge the limitations of their research and provide thorough and exemplary solutions to problems. One concern that has not been raised is the question of whether the study is merely presenting historical facts already known about the rule of queens in an econometric study. However, it is clear that this paper does far more than examine if queen ruled polities are more likely to be involved in conflict. In closing, the authors justifiably note the findings from this paper are not directly applicable to modern-day leaders, and the idea that female leaders would currently be more aggressive in terms of conflict participation than male leaders is not a finding from the paper.

[1] A polity is an organised political entity or community, characterised by a system of governance and institutions regulating public affairs.