How do civilians prevent their militaries from engaging in politics? Scholars are divided in their answer to this question, with some highlighting the constraining effects of political institutions and others emphasizing the importance of norms of civilian control. We integrate these two approaches and develop a dynamic theory of civilian control, arguing that control over the military is achieved once civilianized institutions are adopted and sufficient time has passed to permit: (1) the development of a shared norm of civilian control within the military; and (2) learning among military elites that fosters a belief that civilian rule is robust to military challenges. As a result, civilian control is self-reinforcing. We evaluate these claims by developing a latent variable model that tests for the presence of self-reinforcing institutional dynamics. We generate estimates of civilian control for all countries, 1945-2010 and find strong support for our expectations. In the summer of 2016, Turkish military personnel seized control of the Bosphorus bridges as jets flew overhead in Ankara. Members of the armed forces were attempting to oust President Recep Tayyip Erdŏgan from office. Although it quickly failed, the coup attempt took many by surprise. The years of rampant military intervention in politics appeared to have been over in Turkey. The previous two decades were characterized by stable civilian rule, enough time to suggest that the democratic regime should have consolidated itself from such challenges (Huntington, 1993; Svolik, 2015). The conventional wisdom was that Erdŏgan successfully defanged the military as a source of political opposition (Hannah, 2016; The Economist, 2016). Yet, as these events demonstrate, the military is seldom tamed easily. Why were observers so surprised by the Turkish coup attempt and why does control of the military remain tenuous in regimes where seemingly strong civilian institutions have been established? These questions are central to our understanding of civilian control, defined here in terms of the extent to which civilians dominate political decision-making within a polity, and the robustness of this dominance to military involvement in politics.1 Existing answers to this question broadly originate from two theoretical frameworks. The first is institutionalist and understands civilian control as an outcome behavior emerging from civilian and military elites pursuing unique interests within a fixed institutional setting. Civilian control is achieved when the structure of political institutions is such that the preferences of civilian elites dominate those of the military. This approach links civilian control with political regimes where civilians enjoy a monopoly on de facto, if not de jure authority over members of the armed forces.2 While this provides a robust and generalizable means of understanding civilian control across polities, it ignores the fact that 1For definitions of civilian control relating to civil-military preference divergence and policy implementation and compliance see Feaver (1999, 2003), and Desch (1999). 2 For qualitative examples of institutional typologies of civilian control see Nordlinger (1977); Stepan (1988); Welch (1976); and Colton (1979). The quantitative literature has employed a variety of regime type indicators and indexes to capture this construct (Poe and Tate, 1994; Lai and Slater, 2006; Sechser, 2004; Weeks, 2008, 2012; Talmadge, 2015). Game theoretic treatments of civil-military interactions as a single-iteration game also assume a fixed institutional setting, though recent work has begun to incorporate temporal dynamics into formal models of military coups (Little, 2016).

中文翻译：

---

自我加强的平民控制：基于计量的军民关系分析

平民如何防止其军队参政？对于这个问题，学者们意见分歧，有的学者强调政治机构的约束作用，有的学者强调平民控制规范的重要性。我们将这两种方法结合起来，并发展出一种动态的平民控制理论，认为一旦采用平民化机构并经过足够的时间就可以实现对军队的控制，则：（1）在军队内部制定共同的平民控制规范; （2）在军事精英中学习，使人们相信，平民统治对军事挑战是有力的。结果，平民控制是自我强化的。我们通过开发潜在变量模型来评估这些主张，该模型测试自我强化制度动力的存在。我们估算了1945-2010年间所有国家的平民控制情况，并为我们的期望提供了有力支持。2016年夏天，随着喷气式飞机在安卡拉上空飞行，土耳其军事人员夺取了对博斯普鲁斯海峡大桥的控制权。武装部队成员正试图罢免总统雷杰普·塔伊普·埃尔登。尽管很快失败，但政变企图使很多人感到惊讶。军事上对政治进行广泛干预的岁月似乎已经结束。前二十年的特点是稳定的平民统治，有足够的时间表明民主政体应该已经从这些挑战中巩固了自己的地位（Huntington，1993； Svolik，2015）。传统的观点认为，埃尔多安成功地将军队贬为政治反对派（Hannah，2016；《经济学人》，2016）。然而，正如这些事件所表明的那样，军队很少被驯服。为什么观察员对土耳其的政变企图感到如此惊讶，为什么在建立了看似强大的民政机构的政权中，对军队的控制仍然微不足道？这些问题是我们对平民控制的理解的核心，这里的定义是，平民在政治内部占主导地位的政治决策的程度，以及这种对军事参与政治的统治地位的鲁棒性。1这个问题的现有答案大体上源自从两个理论框架。第一个是制度主义者，他将平民控制理解为在固定的制度环境中追求独特利益的平民和军事精英所产生的结果行为。当政治机构的结构使得平民精英的偏好支配着军队的偏好时，便实现了平民控制。这种方法将平民控制与政治制度联系起来，在这种制度中，平民对武装部队的成员享有事实上的，甚至不是法律上的权力的垄断权。2尽管这提供了一种了解跨政界的平民控制的强大而可概括的手段，但它忽略了以下事实： 1关于与军民偏好分歧，政策执行和遵守有关的民用控制定义，请参见Feaver（1999，2003）和Desch（1999）。2有关民用控制制度类型的定性例子，请参见Nordlinger（1977）；斯蒂芬（1988）; 韦尔奇（1976）；和科尔顿（1979）。定量文献采用了多种政体类型指标和指标来捕获这种结构（Poe和Tate，1994； Lai和Slater，2006； Sechser，2004； Weeks，2008，2012； Talmadge，2015）。尽管最近的工作已开始将时间动态纳入军事政变的正式模型中，但将军民互动的博弈论方法作为一个单迭代的游戏也假定具有固定的制度设置（Little，2016）。Sechser，2004年。星期，2008年，2012年；Talmadge，2015年）。尽管最近的工作已开始将时间动态纳入军事政变的正式模型中，但将军民互动的博弈论方法作为一个单迭代的游戏也假定具有固定的制度设置（Little，2016）。Sechser，2004年。星期，2008年，2012年；Talmadge，2015年）。尽管最近的工作已开始将时间动态纳入军事政变的正式模型中，但将军民互动的博弈论方法作为一个单迭代的游戏也假定具有固定的制度设置（Little，2016）。