This week, I was most intrigued by Lake and Baum’s piece “The Invisible Hand of Democracy,” probably because I would really like to believe what they conclude in their paper. Prior to the publication of their work in 2001, there was generally no robust empirical support for the normative idea that citizens living in a democracy are better off than those in a dictatorship. These authors’ goal was to fill this identifiable gap in empirical evidence to somehow provide proof that democracy has real and substantively important effects on the daily lives and well-beings of individuals. Beyond this, they believe that this positive effect of democracy can be achieved fairly quickly; within a few years of a transition to democracy, one should expect the health and wellbeing of the people to improve.

The author’s theory builds on a class of public choice models, wherein the state is conceived of as having a monopoly on the legitimate use of force, which gives them a natural monopoly within their core areas of production. However, states are still contestable, and can be displaced more or less easily depending on the barriers that exist prohibiting potential competitors, as well as the cost of political participation. Accordingly, in democracies, where these barriers are low, “the state will produce as a regulated monopoly, provide relatively larger quantities of goods at relatively lower prices and thereby earn fewer supernormal profits or monopoly rents.” Conversely. in autocracies, where these barriers are high, “the state will exercise its monopoly power, provide fewer public searches, and earn greater rents. In sum, their hypothesis is as follows: in democracies, states will earn fewer monopoly rents and produce a higher level of services than autocracies.”

This hypothesis is tested both cross-sectionally and over time for a variety of ‘quality of life’ indicators, i.e. public health and education. The results of their statistical analysis confirm their hypotheses, allowing Lake and Baum to substantively conclude that democracy does indeed have real, important, positive effects on the well-beings of individuals regardless of the region in the world.

At face value, this week’s readings seem examine a relationship which I would have described as self-evident, logical, obvious… a ‘given.’ The general consensus has always been that democracies are less repressive and less likely to violate human right than their authoritarian counterparts; the question that remains is, thus, whether or not this relationship is a causal one or simply correlative. Neither the Zanger nor the Davenport and Armstrong pieces fully address this question, but they do both try to explicate the nature of the correlation quite well, the first with specific focus on transitional regimes,  and the latter aggregating different ‘democracy’ measures in order to examine the relationship at different ‘levels’ of democracy.

Zanger’s approach is unique; in trying to better understand the correlation between regime type and repression she looks at the effect of regime change on life integrity violations. In this way, she addresses the question of correlation vs. causation better than Davenport and Armstrong do. There are, however, a number of methodological weaknesses with her analysis. One concern I had was with her ‘anocracy’ classification of regimes. It was very loosely defined as an ‘incoherent regime;’ basically anything that was not a democracy or an autocracy. The lack of a rigorous definition translates to her operationalization of the ‘regime change’ concept. Using the Polity III dataset, she gives the classification of ‘anocracy’ to all regimes receiving scores between -3 and +3, but provides no substantial justification for this arbitrary grouping arrangement.

Davenport and Armstrong’s article provided more robust findings. They were concerned with explicating the relationship between democracy and domestic peace, which is often characterized as negative and linear. Specifically, they wanted to see if the ‘level’ of democracy affected the level of repression within a state. I found their results quite compelling; they found that a threshold model best explained the relationship between democracy and domestic peace. This model posits that “below a certain level, democracy has no effect on human rights violations, but above this level democracy influences repression in a negative and roughly linear manner” (Davenport and Armstrong 2004: 538).

Assessing the legitimacy of the democratic peace theory is not an examination of whether or not there exists a relationship between the dependent variable, dyadic democratic peace, and the independent variable, dyadic democratic governance. There is a long-established empirical pattern supporting the significantly lower likelihood two democracies engaging in militarized interstate disputes when compared to their autocratic or mixed dyad counterparts. Letzkian and Souva 2009 provide one such iteration of the patterns posited by democratic peace theory, by comprehensively examining data from 1946 to 2000.  While the robustness of the pattern is easily discernable, whether or not this pattern is actually indicative of a correlative or causal relationship has been a matter subject to heavy scrutiny. O’Neal and Russett provide an interesting critique of the relationship between democracy and international peace.

To set the foundations for the critique of O’Neal and Russett’s “The Kantian Peace,” a definition of what exactly the “Kantian Peace” entails must first be established. Kant’s classic argument states that an international perpetual peace can exist if the following features are present: a republican constitution, a federation of interdependent republics, and a commercial spirit. Thus, Kant is not proposing a unilateral relationship with a single independent to dependent variable association; the requirements for Kantian peace necessitate the examination of at least three other independent variables. Accordingly, O’Neal and Russett aim to assess (1) democracy, (2) trade interdependence and (3) IGO involvement, in relation to peace. By performing both dyadic and systemic analyses, O’Neal and Russett ultimately conclude that Kant’s theory was remarkably prescient.

To assess the verity of O’Neal and Russett’s hypotheses, their variables and their respective measures must first be overviewed. They use a single dependent variable, DISPUTE, a dichotomous measure of whether or not there is a militarized interstate dispute (MID), and three Kantian dyadic independent variables: DEM_L, a gradated measure of the level of democracy in the less democratic state of each dyad, DEPEND_L, a gradated measure of the level of dependence of the state less economically dependent on trade, and IGO, a gradated measure of the number of IGOs in which both states in a dyad share membership.

To best dispel the possibility of reverse causality for their purposes, O’Neal and Russett lag all independent variables from the dependent variable by one year. This precaution aims to protect again endogeneity i.e. when conflict may limit trade just as trade may constrain conflict. They also try to control for a variety of realist explanations for dyadic peace, including in their analysis the capability ratio (CAPRATIO), alliance (ALLIES), and contiguity and distance (NONCONTIG and DISTANCE). However, they do not limit the dyads examined to those that are ‘politically relevant,’ a decision which is justified as an attempt to make sure no patterns are overlooked as a result of exclusion. Ergo, in their analysis O’Neal and Russett will also give the overall dyadic results separately from ‘politically relevant’ dyadic results, and compare the two. For the years 1885- 1992, excluding the two world wars and the immediate postwar years (1915-20, 1940-46, O’Neal and Russett observed a total of 150,000 dyads.

To assess their observations, O’Neal and Russett employ a logistic regression analysis. A method of weak-link specification is applied for the DEM_L and DEPEND_L variables, but the IGO variable is inherently dyadic. O’Neal and Russett also assess the realist control variable, to provide the following equation in summation:

(EQ₁) DISPUTE= DEM_L + DEPEND_L + IGO + ALLIES + CAPRATIO + NONCONTIG + DISTANCE + MINORPWRS

As this paper is primarily concerned with the relationship between democracy and peace, the first row of Table 1 is requires the most attention. When considering all dyads in model 1, a higher DEM_L is statistically significant at the .001 level, to a greater likelihood of peace. To test if this relationship exists outside of the cold war era, GEE is used again to re-estimate EQ1 for the years from 1886-1914 and 1921-39 (model 3), and the DEM_L remains significant at the .001 level. To investigate the possible discrepancies between including all possible pairs of states and ‘politically relevant’ dyads, EQ1 is re-estimated a third time, using only ‘politically relevant’ pairs of states, or those that contain at least one major power and those which are contiguous (model 4). Here, the positive relationship between democracy and peace remains apparent (p < .001). The results of this analysis lead to O’Neal and Russett’s conclusions that, compared with the typical dyad, the risk that a more democratic dyad will become engaged in a dispute is reduced by 36 percent, and the same risk for a more autocratic dyad is increased by 56 percent.

One of the other Kantian variables examined in O’Neal and Russett’s piece, trade interdependence (DEPEND_L), takes on especial importance for the purposes of its subsequent comparison with Gartzke’s piece. The GEE estimations posit that the relationship between trade dependence and peace is equally as statistically significant (p < .001) as the relationship between democracy and peace for all dyads for all years[2], and for the pre-Cold War era, and almost as significant (p < .002) for politically relevant dyads[3]. The correlations in realist variables also play a role in the upcoming discussion of Gartzke, with O’Neal and Russett’s results corresponding to all expected patterns:

(1) A preponderance of power rather than a balance deters conflict;

(2) Contiguous states are prone to fight, as are those whose homelands are geographically proximate; and

(3) Major powers are involved in disputes more than are smaller states.

All of these realist variables are significant at the .001 level.  (51)

It is clear from the readings this week that there is an empirically observable relationship between democracy and economic growth. What remains unclear is whether or not this relationship is a causal one. Gerring et al’s piece attempts to establish a number of causal mechanisms linking democracy to economic growth, with the important scope condition of time. The following figure, taken from their article, outlines the main causal mechanisms identified:

The discussion they present of democracies’ creation of physical, human, and social capital do not introduce any new or original ideas. In their discussion of political capital, however, the time factor becomes a subject of emphasis. Democratic political arrangements create a number of political outputs that are assumed to have a direct impact on economic performance, i.e. “market-augmenting economic policies, political stability (understood as a reduction of uncertainty), rule of law, and efficient public bureaucracies.” Valuations of these aspects of political capital are dependent on regime length, which Gerring et al. describes as a. ‘learning’ and b. ‘institutionalization’ processes. In sum, Gerring et al. conclude that, if maintained over time, democracy influences economic performance through four main channels – physical, human, social and political capital- constituting a ‘democratic growth effect.

Of the assigned readings this week, I found Paxton’s article to be most interest and most empirically compelling. The historical variation in the term ‘universal suffrage’ is not new to any of us. I still recall the political theory course I took in which we discussed the evolution of suffrage from the Ancient Greek, male, non-slave, property owner to what is considered the ‘healthy democratic’ standard today (anyone of ‘voting’ age). Paxton focuses particularly on the inclusion (or, rather, exclusion) of women and the misappropriation of the term ‘universal suffrage.’ In doing so, she finds fault with a number of previous measures of democracy, outlining the discrepancies between the definition and measurement of democracy, and the resulting theoretical and practical consequences.

Interestingly enough, I have read this piece before, but for its contributions to a feminist strand of political thought. Revisiting it as a critique of measurement flaws, instead, has actually given me a greater appreciation of Paxton’s work. She thoroughly and bravely explicates the misalignment between the operationalization and conceptualization of suffrage and, consequently, democracy in some of the most seminal works in political science literature (i.e. Huntington (1991), Rueschemeyer, Stephens and Stephens (1992), Bollen (1980), Lipset (1959)). Paxton’s own argument is a compelling one: “the decision to exclude women’s suffrage from measurements of democracy can affect three areas of research: (1) scoring dates of democratic transitions, (2) descriptions of the emergence of democracy, and (3) research on understading the causes of democratization.

The suggestion she provides to address this problem relate back to our discussions of dichotomous vs. gradated measures of democracy. She notes that dire oversights like the exclusion of women in ‘universal suffrage,’ are more likely to occur when democracy is measured as a dichotomously, or even categorically. Thus, Paxton concludes that, in order to capture a fuller view of democracy, it should be measured on a gradated scale.

Coppedge and Reinicke present a detour to the realm political science literature dedicated to the conceptualization and measurement of ‘democracy’ by, instead, measuring ‘polyarchy,’ in strict accordance with Dahl’s eight institutional requirements. In Dahlian theory, ‘democracy’ is an intrinsically unattainable ideal type of government and ‘polyarchy’ is the closest concrete approximation of it. However, for Coppedge and Reinicke, this difference is merely a matter of semantics (consider the similarities between the variables they study and the variables studied in various other measures of ‘democracy’); thus, their measure becomes most clearly differentiable from the others studied via their usage of Guttman scaling, which creates an ordinal scale placing countries  into ranked categories. The application of Guttman scaling to their results eliminates a quintessential concern in quantitative political science research: how to balance the weighting of the individual components of any given measure; on the polyarchy scale, none of the 4 variables are treated as more/less important than one another. This approach is not without its caveats. Despite solving the problem of weighting, the use of Guttman scaling also results in a number of countries being grouped at top-tier levels of democracy, detracting from the meaningfulness of the results attained.

Overall, while I did find Coppedge and Reinicke’s approach to be unique and memorable, I am not entirely convinced of its usefulness, especially in the context of the regional democracy report. Furthermore, when taking into account its limitations in temporal scope (with data available only for the years 1985 and 2000), and the number of cases, as few as they may be, that did not fit into a ‘perfect scale type’ there are structural disadvantages to the utility of this dataset as well.

Munck and Verkuilen: “Conceptualziting and Measuring Democracy: Evaluating Alternative indices provides a comprehensive and integrated framework for the analysis of data and uses this framework to assess data sets on democracy. The framework first distinguishes among three challenges that are sequentially addressed: conceptualization, measurement, and aggregation. In turn, it specifies distinct tasks associated with these challenges and the standards of assessment that pertain to each task.

This framework is applied to the data sets on democracy most frequently used in current statistical research, generating a systematic evaluation of these data sets. The authors’ conclusion is that constructors of democracy indices tend to be quite self-conscious about methodological issues but that even the best indices suffer from important weaknesses. More constructively, the article’s assessment of existing data sets on democracy identifies distinct areas in which attempts to improve the quality of data on democracy might fruitfully be focused.

I found Adcock and Collier’s piece “Measurement Validity: A shared standard for qualitative and quantitative research to be the most worthy of note this week. Their task, in and of itself, is quite unique. Rarely are political science articles so loose in focus; the topic of measurement validity is such a vast and diverse one. Often, papers will be considered with simply critiquing the critique of one other paper’s particular operationalizing of a concept, and its validity, not the entire issue of validity en masse. Adcock and Collier confront this daunting task by taking a thematic approach. First, they seek to “establish a shared framework that allows qualitative and quantitative scholars to assess more effectively, and communicate about issues, of valid measurement. This deals directly with the aforementioned instances of one author critiquing another author’s work based on its validity. Adcock and Collier’s piece seeks to establish some kind of common ground that will circumvent the fruitless back and forth that often occurs.

Secondly, Adcock and Collier bring to the forefront of the discussion the need to draw a clear distinction between measurement issues and disputes about concepts. Again, this speaks to the perpetual back-and-forth between political scientists, by suggesting that, in some cases, there may exist an intrinsic disagreement between authors regarding the conceptualization of any given concept, as opposed to the measurement of it.  Thirdly, the authors capitalize on the conceptual specificity of measurement claims, and suggest that claims can be more generalized without losing their validity oif greater attention is paid to matters of context. Finally, Adcock and Collier address alternative measurement validation procedures and give three examples of types of measurement validation, namely, (1) content validation), (2) convergent/ discriminant validation, and (3) nomological/ construct validation.

They assess each as serving different purposes, and emphasize the usefulness in differentiating between these three types of validation. Adcock and Collier sought to present an “understanding of measurement validation that can plausibly be applied in both quantitative and qualitative research.” They achieved this purpose by reframing each type of validation so as to create basic questions that can be utilized in both quantitative and qualitative analysis. In this way, they are hoping that qualitative and quantitative researchers can resolve their differences and learn from one another.

In Collier and Adcock’s Democracy and Dichotomies: A Pragmatic Approach to Choices about Concepts, they present a pragmatic peace offering to the longstanding disagreement regarding whether a dichotomous or a gradated approach to the distinction between democracy and nondemocracy is more appropriate.

In doing so, they do not advocate one approach in favor of another, but, instead, suggest that the choice between the two should be made on a context-specific basis, depending on the concepts being operationalized in any given measure of democracy. They do so by examining the conceptual justifications motivating various scholarly proponents of the two opposing approaches at different levels of generality; the general methodological defense, followed by generic justifications, and concluded by context-specific justifications. Collier and Adcock find the last to be the most convincing; in research that focuses on democratizations as a well-bounded event, as well as on classical subtypes of democracy, they see dichotomous approaches as being preferential; however, they acknowledge that alternative methods of evaluating democratizations and subtypes call for gradated measurements. Other contextual considerations, such as normative concerns and the empirical distribution of cases, also contribute to the decision between dichotomies and gradations.

Frankly speaking, what Collier and Adcock propose in dealing with the dichotomous/ gradated debate seems self-evident and quite simplistic. Clearly, each approach has its merits depending on what best serves the purposes of the researcher. I do, however, appreciate their discussion of the importance of sharper differentiation points; wherein gradations are combined with named categories. This approach is the one I would personally advocate in classifying regime types; prior to a certain point, no matter how much more ‘democratic’ a country is than a less ‘democratic’ country, it should not be considered as such until a certain bare minimum number of requirements is fulfilled.