jandresgannon
commented
4 years ago A template for the model section from Bleek, Philipp C., and Eric B. Lorber. “Security Guarantees and Allied Nuclear Proliferation.” Journal of Conflict Resolution 58, no. 3 (April 1, 2014): 429–54. https://doi.org/10.1177/0022002713509050.
4 Quant empirical analysis 4.1 Building off existing datasets Most recent data builds on X or Y datasets. We use X because broader temporal scope. The structure of the data is time-series cross-sectional with the unit of analysis being country (monad) year
4.2 Who did what when? How we code the DV What's the time period We code behavior for the DV into the following categories defined this way
4.3 Operationalizing security guarantees Recent work has operationalized the EV in the same way, but we do something new We create 2 variables to capture X and Y We identify this using the COW alliance dataset because X
4.4 Controlling for other factors "we first construct ‘‘core’’ models that include only a small number of potentially confounding control variables, those which theory suggests are causally related and prior to both the dependent variable of proliferation and the independent variable of interest, security guarantees. These models are also consistent with recent recommendations of methodologists skeptical of so-called garbage can approaches (Ray 2003; Achen 2005). At the same time, the studies on which we build and which we critique control for a broader range of factors, and, while our primary goal is to shed light on how security guarantees affect proliferation, we also aim to explain proliferation broadly. A second set of ‘‘fully specified’’ models includes a larger number of variables, some of which are less tightly linked to the independent variable of security guarantees. We chose these variables by starting with all fourteen factors for which Jo and Gartzke controlled in their 2007 study, minus two they employed to account for temporal dynamics, since temporal dynamics are internal to our hazard modeling approach. To these twelve, we added four for which other scholars have made robust theoretical cases, but found that sixteen independent variables reduced significance across the board. We therefore dropped Jo and Gartzke’s domestic unrest and diplomatic isolation variables, which enjoy little theoretical support and were never statistically significant in either our models or theirs, leaving us with the fourteen independent variable model we present in the following. All variables are drawn from Jo and Gartzke unless otherwise indicated. Our ‘‘core’’ models include four control variables. Since the acquisition and possession of nuclear weapons incurs significant economic cost, states’ ability, and therefore motivation, to proliferate should be influenced by the availability of economic resources (Meyer 1984). States’ economic resources should also affect their ability to enter into formal alliances, including with nuclear-armed states, since more well-endowed states will be able to contribute more to the alliance, other states will have more interest in protecting them, and economic resources buy leverage states can employ to induce others to protect them (Stein 1990). Our economic capacity variable is a moving average of each state’s fraction of global energy consumption and iron and steel production. Proliferation is also technically challenging. States with more substantial nuclear material and technological resources at their disposal should therefore be more able to proliferate (Hymans 2006; Jo and Gartzke 2007). Because such states pose greater proliferation risks, nuclear-armed patrons should be more willing to extend security guarantees to them to reduce the likelihood they will proliferate (Oberdorfer 2001). Our latent nuclear weapons production capacity variable controls for nuclear infrastructure and resources, summing ‘‘the number of resources or production capacities that a given state has for nuclear weapons production. The resources and capacities consist of the seven components (uranium deposits, metallurgists, chemical engineers, and nuclear engineers/physicists/chemists, electronic/explosive specialists, nitric acid production capacity, and electricity production capacity)’’ (Jo and Gartzke 2007). At least some of these are likely endogenous to the decision to pursue nuclear weapons—for example, states may prospect for uranium in part because they harbor weapons ambitions—but as a robustness check, substituting a rudimentary industrial capacity threshold variable yields similar results. States may pursue nuclear weapons in the face of severe conventional security threats (Sagan 1996–1997). States may also seek and receive security guarantees in order to try to ameliorate such threats (Goldstein 2000). Our conventional military threats variable is equal to the summed capabilities of a state’s enduring rivals divided by that state’s capabilities, with one added and natural log transformed. Capabilities are drawn from the COW’s Composite Index of National Capabilities. Relatedly, states may pursue nuclear weapons—or seek security guarantees—in the face of nuclear security threats by an adversary (Goldstein 2000). A variable capturing whether one or more enduring rivals have acquired nuclear weapons is imported from Bleek (2010). 15 Our ‘‘fully specified’’ models include a broader set of control variables. Each of these variables is robustly theoretically linked to proliferation, but in some cases less robustly to security guarantees. Kroenig (2009b, 2010) argued that whether states receive sensitive nuclear assistance is key to explaining why states acquire nuclear weapons, because such assistance helps states overcome technical barriers to proliferation. This argument could be extended to launching nuclear weapons programs or even exploring nuclear options. While a weaker argument, states might be more likely to receive guarantees from nuclear-armed patrons willing to share sensitive nuclear materials or technology, or patrons might be more willing to extend guarantees to sensitive nuclear assistance recipients because they fear their proliferation. We import from Kroenig (2009b) a variable capturing whether states receive sensitive nuclear assistance, that is, uranium enrichment, plutonium reprocessing, or direct weapons assistance. Fuhrmann (2009) argued that whether states receive civil nuclear assistance is a key determinant of both pursuit and acquisition of nuclear weapons, because such assistance develops capabilities on which states can build weapons programs. The relationship between civil assistance and security guarantees is less intuitive, though as with sensitive assistance, it is possible that states providing civil assistance may be more willing to extend security guarantees or that the capabilities civil assistance provides may make potential patrons fear proliferation and therefore be willing to extend guarantees. We import from Kroenig (2009b) a variable capturing whether states receive foreign assistance constructing research or power reactors. We regard this variable as superior to one employed by Fuhrmann, which counts nuclear cooperation agreements signed. 16 As a robustness check, we reran our analyses using Fuhrmann’s variable and observed no changes to our reported results. The 1970 Nuclear Non-Proliferation Treaty (NPT) is widely cited as having stemmed proliferation by reassuring states that their neighbors and rivals who joined as nonnuclear weapon states were not proliferating and that nuclear-armed states were working toward marginalizing and eventually abolishing nuclear weapons (Scheinman 1990). Whether states have joined the NPT, and more broadly its perceived robustness, may also affect whether patrons extend formal, bilateral security guarantees, though that theoretical case is less compelling. We capture the potential effects of the NPT in two ways. A variable captures whether states have ratified the NPT. A related variable, equal to the proportion of states in the international system that have ratified the treaty in any given year, controls for the existence and strength of the NPT. Because of the role major powers play on the world stage, both status and functional concerns may drive them to proliferate (Jo and Gartzke 2007). Similarly, those major powers who have not obtained nuclear weapons, such as Japan, may be likely to seek security guarantees from nuclear-armed patrons. Our variable relies on the standard COW coding of major powers in international politics. 17 Similar to major powers, states that play major roles in their regions may consequently desire nuclear weapons, even independent of their greater capabilities to proliferate and possibly more threatening security environments (Jo and Gartzke 2007). And if they are unable or choose not to proliferate, regional powers may be more likely to seek security guarantees. Regional powers are identified as those whose aggregate capabilities are equal to at least half those of the most powerful states in their regions and who are not major powers. Democracies may be more or less likely to proliferate. Some argue autocracies are better positioned to launch and sustain secretive weapons programs, others that populist politicians may pursue weapons to stoke nationalist desires (Singh and Way 2004; Jo and Gartkze 2007). Democracies may also be more likely recipients of security guarantees because they are viewed as more reliable and pacific alliance partners (Doyle 1986). A variable measures the level of democracy for each state in each year, by integrating rankings for both democracy and autocracy from the Polity IV data. Solingen (1994, 2007) has argued that states’ integration with the international economy affects their proliferation proclivity, because ruling coalitions choosing more inward-oriented strategies face fewer costs and greater rewards from prolifera tion, while those choosing integration face greater costs and fewer rewards. It is plausible that more integrated states might also be more likely to seek and receive security guarantees; Solingen (1994) makes a version of this argument with special reference to the Japanese, South Korean, and Taiwanese cases. We import from Singh and Way variables measuring openness to the international economy and changes in openness over time."
4.5 Modeling approach We use model X. We make modeling decision Y based on criteria Z
4.6 Multivariate analysis Here are the models and results EV is statistically significant and negative Other results are unsurprising Here's the control results
List our current controls and compare to recent JCR models. Then re-estimate our models with those controls added on.