Week 3. February 9. Institutions, Endogeneity and Path Dependence
Post your comments/questions as comments here-- feel free to post under your Google profile or 'anonymously,' but in either case sign your first name in the post so we know who's making each comment.
I argue that Martin Shefter's detailed historical analysis suffers from its small sample size. Selection bias is always a potential problem in these kinds of studies, but in this instance I think that the small sample size has allowed Shefter to obfuscate with respect to several ways of explaining variation in party and patronage across countries.
Firstly, he steadfastly refutes the neoclassical theory of patronage at the outset of his paper. He does this by citing several important counter-examples. However, at the end of the essay, when acknowledging the existence of important counter-examples to his own argument, he argues that his theory need not ‘utterly predetermine the subsequent character of a nation’s party politics’. Given this, is it wise to be so certain that the more ‘sociological’ explanations offered by the neoclassical theory do not also play a role?
Secondly, he offers little room for political agency in his account. Political actors who may benefit from systems of patronage may seek to find ways to supply offices and rents to supporters while maintaining the veneer of an independent bureaucracy or placating such (liberal or absolutist) forces as oppose patronage. The layers of special offices and positions created by the United States President with discretionary appointments are examples of this. In Ireland, similarly, an independent civil service existed at the time of independence in the early 20th century. The civil-service, largely based upon the British model, was retained in form and organization by successive Irish governments. This did not stop, however, successive governments from building layers of special commissions, advisors, officers, councils and the like on top of and alongside the regular civil service. These former positions provided the means for political patronage. Shefter’s discussion does not, I think, leave sufficient room for such hybridized cases.
Acemoglu, Johnson and Robinson’s use of mortality rate as an instrument for current institutions leads to interesting conclusions regarding the choices of colonial institutions and the persistence of those institutions throughout history. According to their conclusion, the influence of mortality rates—determined by condition in colonies—on colonizers’ decision to settle or exploit, and thus their choice of institutions to implement in the colonies, still exerts effect on economic development. This finding supports the belief that political institutions—particularly those of property rights—cause economic growth, and not the reverse.
For Acemoglu et al., institutions persist. In fact, the lasting effect of extractive versus neo-European institutions has persisted through independence, across borders, and through various stages of state formation—be those states under democratic or authoritarian rule, post-independence. Their argument that institutions persist is not convincing, particularly when considering that many states included in their observations underwent institutional changes throughout their history. In particular, when contrasting this argument to their “Reversal of Fortune” article, published a year after “Colonial Origins,” it is exactly the shift in institutions that causes the reversal of fortune. If institutions persisted, then it would be difficult to convey an “institutional reversal.” However, colonial intervention created a disruptive shift, so that societies benefitting from higher levels of income and urbanization were left with extractive institutions, while in previously poor countries, institutions were created to foster investment. If we can consider the possibility that colonization can exert a strong shift, we can also consider the possibility that independence or a change in regime can exert a similar effect on institutions, so that institutions may not necessarily persist throughout these experiences.
Furthermore, the use of mortality rates as an instrumental variable may not accurately capture colonizers’ choices. Referring to Appendix Table 2, we see that Panama had a high settler mortality rate. However, as opposed to enacting extractive institutions, Spanish settled in the area, using Panama as a control base for neighboring colonies. On the other hand, South Africa, which displayed settler mortality rates close to that of the U.S. (15.5 versus the U.S.’s 15 and Canada’s 16.1), was colonized with extractive institutions, particularly after the discovery of diamond and gold in the region. Engerman and Sokoloff’s argument for the relevance of factor endowments in determining colonial institutions seems to be more appropriate, under this scenario.
One of the issues that I found especially interesting in this week’s readings is how to effectively measure institutions. Glaeser et al. criticize the three sets of measurements that have been commonly used in existing literature as “measuring only outcomes… not reflecting either constrains on government or permanent features of the political landscape”. Instead, they propose a more “deep” and precise measurement using electoral rules, judicial independence and constitutional review that are claimed to reflect permanent characteristics of institutions. Both measurements, the existing ones and the newly proposed one, are flawed in my opinion. Although the former ones that are constructed by survey data on perception of institutional quality and government effectiveness are comprehensive, they are inevitably contingent on interviewees’ cognitive ability and subjective bias; likewise, the way of selecting some indicators to proximate institutions is problematic because the selection per se is subjective and impossible to exhaust all “permanent” characteristics of institutions. The measurement problems of institutions can be attributable to the ambiguous definitions of institution and its extension. Take the dictator’s choices mentioned in Glaeser et al’s essay for example, should the dictators’ self-disciplining in expropriation and property rights violation as well as their developmental rationale be counted as a benign institutional constrain for economic development in some countries? The answer really depends on how we define “informal institutional constrain”. All of the institution measurements mentioned in Glaeser et al’s article are deficient in capturing some informal yet substantial institutional constrains in the reality.
In my mind, Acemoglu, Robinson, and Johnson’s (ARJ, 2001) original finding and Glaeser et al.’s (2004) alternative are both incomplete arguments, because argue deep institutional explanations based on the constraining power of initial conditions are key to development outcomes without offering even a cursory reason as to why the externalities imposed by such initial conditions should continue to hold over time. In effect, they use Hall’s (2003) initial path-dependent conception- in that initial conditions influence initial institutional outcomes that persist- without doing enough process tracing to show why institutions get “stuck”. It is too bad that AJR self-consciously leave institutions and the mechanism by which they persist as a black box, but Glaeser et al. spend a bit of time discussing the externality feedbacks generated by human capital, on the one hand, and institutions on the other. Given this, it is too bad that they don’t expand this a bit to show how initial human capital endowments shape incentives for favoring particular types of institutions that generate a feedback loop and allows both good institutions and high human capital to persist over time.
Further study of both of these theories is a great opportunity for future work, because although institutions are widely regarded as sticky in the literature, an understanding of why and how this process occurs has been lost under broad generalities about “political culture”, “social mores”, and underspecified externalities. The sorts of deep institutional questions noted above allow for unique opportunities to disaggregate parametric requirements (a la Grief and Laitin, 2004) for equilibrea in order to get a better grasp of which factors tend to be more central to strategic choices at particular times by allowing for long observations. For example, in the case of Glaeser et al., where the key determinant of such parameters is posited to be human capital, it should be possible to craft a research plan centered around understanding how changes in human capital endowments may have shifted incentives for particular types of institutions and, more interestingly, why such shifts did not occur uniformly with increasing human capital endowments.
The discussion over the relationship between economic growth and institutions advanced by Acemoglu, Johnson and Robinson (AJR), as well as Glaeser et al. is a perfect example of the frustrating nature of the "chicken or egg" debates that dominate this issue. Both papers provide compelling empirical evidence for their findings, but their conclusions are contradictory. By pointing out, for example, glaring problems with the traditional measurements of institutions (i.e., confounding institutions with their outcomes), Glaeser et al. show that, even if AJR’s methodology is sound (which itself is questionable), the chosen variables are biased. This conversation demonstrates the ambiguity and difficulties associated with empirical analysis of social phenomena. That said, if scholars choose this particular type of variable-based methodology, it seems that the discourse will simply descend into a never-ending commentary on such important, yet inconclusive and tangential issues as variable selection and measurement bias.
Hall provides a refreshing perspective on this dilemma, as well as a convincing strategy for resolving it. Since interactions and dynamic processes make social phenomena too complex for traditional regression analysis, political science scholars should focus on devising a methodology that "catches up" to the current ontology that incorporates strategic decision-making and time elements. It seems persuasive to me that focusing on detailed, small-N studies can yield valuable information about processes that would enhance discussions such as that between AJR and Glaeser, et al. In fact, as other scientists already recognize, detailed examinations of particular cases are bound to reveal consequential causal information that would otherwise be assumed away or overlooked in traditional macro-level regression analysis. Furthermore, it would wean political science scholars off the tendency to ascribe more causal weight to statistical inference than traditional regression models can bear.
Glaeser et al. argue that three of the measures of institutions from the economic growth literature do not, in fact, measure institutions. Their argument creates doubts about the proposal that economic development can be attributed to institutions. One of the measures they highlight is the risk of expropriation used by Acemoglu, Johnson, and Robinson. Glaeser et al. argue that this measure, along with a measure of government effectiveness, measures outcomes instead of institutions. Their definition of institution requires that they are permanent or durable constraints on government. AJR's measure is neither a constraint (as it includes the free choices of dictators) nor permanent (it rises with income and is highly volatile). Glaeser et al. propose an alternative measure of institution, one that does not look at assessments of institutional outcomes, but rather manipulable rules and procedures. While their suggestion would move the measure closer to their definition of institution, it does not fully explain why institutions are persistent---or why they can change, giving space for an incorporation with Laitin and Grief's analysis.
I appreciate very much Greif and Laitin's purposes in presenting their 'Theory of Endogenous Institutional Change'--both the overall goal of joining the strengths of historical institutionalist analysis and game theory for a more comprehensive approach to institutions, and the more precise goals of creating a framework in which (1)endogenously created 'variables' can act like (exogenous) 'parameters,' and (2) institutions reinforce themselves by widening the range of situations in which the behaviors they engender will act like exogenous parameters to reinforce the institution on its current path. This framework is helpful because it allows an institution to be successful at 'enforcing' the behaviors they entail but be unsuccessful at reinforcing itself. But there are a few points I would like to discuss and have clarification on from any of you who understand this piece well: 1) What, clearly, is the difference between self-enforcement (is this just enforcing the behaviors they dictate/entail?) and self-reinforcement?
2) Do Greif and Laitin 'strawman' historical institutionalist uses of critical junctures? Misuse of this construct seems to be one of their key complaints with historical institutionalism.
3) What does this framework really add to a potential institutional analysis by putting the framework in game theoretical terms? The framework may certainly correct some key mistakes in the historical approach, but much of it seems doable without game theory explicitly. Is this (1) a new and significant contribution, or (2) more simply an attempt to make relevant institutional approaches accessible to the game theory - oriented community in political science, and/or (3) an attempt to make game theory approaches to institutional analysis the dominant ones? (As a bridging exercise between the two broad approaches, it is helpful. But I'm left wondering whether it does much that a clearer, more spelled-out historical institutionalist model could not.) Any ideas?
Both Acemoglu et al. (“The Colonial Origins of Comparative Development) and Glaeser et al. (“Do Institutions Cause Growth”) attempt to use path dependent arguments to account for divergent levels of economic growth in the developing world. The implication of their research project is that historical developments dating back to the colonial period have had a longstanding impact on current levels of development in some of the world’s poorest countries. Yet, both of their articles seemed to “miss the mark” in terms of providing a compelling explanation of how their independent variable of interest (“institutions” and “human capital”) are causally related to economic growth. In the case of these two pieces, it seems as though another research method, perhaps one involving a more robust discussion of historical developments through process tracing would have been more appropriate. Both of the authors’ insistence on validating their argument using statistical tests resulted in the neglect of any discussion of the complex political and economic processes underlying the correlation that they identify in their respective studies. These approaches are not mutually exclusive. I would argue that the use of both would have made the arguments of the authors much more compelling. Furthermore, it would seem impossible to gain a better understanding of countries that do not fit into their theoretical models without utilizing such an approach. This tendency to propose historical arguments with few or no case studies or comparative analysis is a good example of the disjoint that Hall discusses between ontology, which acknowledges the importance of historical causal factors, and methodology, which is increasingly favoring statistical analysis, in political science.
This weeks readings substantively discuss issues of institutional change – the how, why, when, consequences of, and rationale for shifts in institutional structure. More fundamentally however I was interested in the manner in which a subset of readings deal with issues of research methodology in the political sciences – by either questioning the validity of traditional constructs in statistical methods (Glaeser et al. 2004; Acemoglu et al 2001); exploring the need for alternative approaches that balance rational and structural methods (Kitschelt 2002; Hall 2003); or improving and refining existing concepts of path dependence (Page 2006; Laitin and Grief 2004). As a newcomer to political science it was interesting to see the debate on institutions remain so centered on issues of process, and the degree of ambiguity that remains. How far has this debate evolved?
Greif and Laitin provide a framework for conceptualizing endogenous institutional change. The concept of quasi-parameter (parameters that over time are changed by institutions and in the long run effect the choices of actors) allows us to think of how institutions can have different short and long term consequences depending on their effects on on these quasi-parameters. It would be interesting to apply their frame work the the Latin American colonial experience described by Engerman and Sokoloff. Although under the rule of the same overall spanish colonial institutions characterized by extreme inequalities, differences in the how authority or grants were distributed among settlers between the different regions could have led to different long term outcome. While those institutions were self-enforcing in that they resulted in economic gains that ensured elites would have an interest in the perpetuating them, the above mentioned differences could have compounded with the growing wealth of elites and led to different outcomes were external parameters to change, for example. Competition among elite groups could lead to the coopting of 'the people' in some scenarios as spanish rule declined, could have undermined the self-reinforcement of the unequal institutions. In general I found Greif and Laitin's theory very useful as it is able to incorporate into an overall game-theoretic framework the very realistic assumption about how institutions, rules, themselves can alter the beliefs and payoffs of different actors over time.
Acemoglu et al. (2001) and Glaeser et al. (2004) posit alternative theories concerning the importance of institutions in explaining economic growth. Each authors’ attempt to offer a “deep” explanation of causal change falls short. Acemoglu et al. fail to provide an historically accurate context for their argument. In the first instance, they try to compare all ex-colonies as if they were settled in the same way in the same time and place. The settlement of Latin America in the 1500s differed dramatically from the way Africa was settled in the 1800s. The world had changed; the Industrial Revolution brought with it new technologies and advances in medicine impacted diseases. Further, institutions in the colonizing nations had changed. Institutions that protected private property were only beginning to surface in Europe during the initial waves of colonization and continued to be perfected over time. Acemoglu et al. also do not specify the criteria they use to differentiate the two types of colonizing types. Indeed, might there be more a graduation in approaches, such as degrees to which institutions were “extracting”? Acemoglu et al. also fail to fully specify how their path dependence argument works – are they making an argument about the importance of path dependency or phat dependency? Likewise, while Glaeser et al. provide a compelling argument, they too underspecify how their theory operates in practice. Indeed, they admit that they do not want to “push [their] results too far,” (293) and that a myriad of other factors could have impacted the colonization process. How is Glaeser et al. a robust theory then if, as they freely admit, their potential instrument term (human capital) may be correlated with the error term? Further, how should the term “human capital” be defined? Glaeser et al. define in primarily in terms of years of schooling. Might there be other variables, such as literacy rates or variables based on skills that may be a better instrument to study in which to measure “human capital”? Glaeser et al. also question Acemoglu et al’s method of institutional measurement. But Glaeser et al.’s measures too suffer from collinearity issues. Further, their proposed measurement (electoral rules, judicial independence, constitutional review) are not as stable in many of the colonized nations as Glaeser et al. assume. Amy Semet
In terms of institutional change and diversity, Herbert (2002) points out the importance of an analysis including both structure and human agency. Avner and Laitin (2004) construct a model which introduces ideas from the game theory and historical institutionalism. Hall (2003) proposes a “systematic process analysis” to cope with the gap between theoretical ontology and methodology. All the above works intend to combine the pros and address the cons of the current research approaches. Yet, none of them provide a satisfactory theory/approach to study institutional change. Herbert simply concludes with the impossibility of prediction. I’m not sure if I’ve learned how to differentiate “variables,” “parameters,” and “quasi-parameters.” Institutions are endogenous. AJR (2001) and Glaeser et al. even show reverse directions of causality regarding the relations between growth and institutions. Since I am incapable of offering a better solution about the study of institutions, I decide to say something about Shefter’s article. It is a historical analysis of group politics. It seems to me that the analysis has less to do with the core assumption of “internally and externally mobilized parties” and their preferences for and against patronage. Besides, with a focus on early group coalition, I think the author neglects the effect of electoral rules on the party’s strategy to grant patronage.
The literature on the effects of institutional design on economic growth has demonstrated that “institutions matter,” but the exact mechanisms by which institutions affect development across countries remain largely unspecified. In AJR (2001), too, the question of which and how institutional processes affected economic (under)development remains unanswered. The causal argument in their paper merely asserts that colonizers established institutions conducive to growth where environmental conditions rendered settlement possible, and that they established extractive institutions otherwise.
But what about cases where colonizers might have encountered both conditions conducive to settlement and high levels of resources? AJR’s settlers would have set up institutions that protected property rights and conferred the benefits of resource extraction to the ruling elite. The initial allocation of economic rents to a small group would have discouraged institutional developments conducive to growth and development, and thus despite property rights, limited access to economic opportunities for the majority of the population would have hindered economic development (Engerman and Sokoloff, 2002). Distributive factors independently affected institutional outcomes and their effects in turn, even when settler migration occurred.
Hence, what AJR summarize in a single causal argument is in fact much more complex. Institutions are shaped and change as a result of geographic, economic, political, and social conditions; analyses of the causal factors of economic development must focus on the micro-processes within countries that take these complex conditions into account. In this context, the effectiveness of broadly focused cross-country regression analyses is limited given that instrumental variables such as settler mortality rates for institutions are often too coarse to permit a close analysis of the causal mechanisms linking institutions and economic development; the invariability of instruments like settler mortality furthermore limits their effectiveness in explaining institutional change.
According to Kitschelt, a good social science explanation has two features: it provides “deep” causes, and at the same time offers microlevel mechanisms that account for the relationship between the independent variable(s) and outcomes. Both features are a call for taking issues of temporality seriously. What comes first and last in determining outcomes (the order of events, or its sequence) is crucial in theory construction in general, and especially important in theories dealing with (endogenous) institutions. How well do explanations dealing with institutions score on both these fronts? Acemoglu et al.’s story goes deep (colonial history) but does not perform well at the level of mechanisms (e.g. how is that old institutions affect new ones?) In contrast, Glaeser et al. explanation is shallow (human capital is too “proximate” a cause of development) and the mechanisms unclear (why do some autocracies choose to invest in education and others not?).
In "Party and Patronage", Martin Shefter seeks to account for international variations in the role played by political patronage. Shefter's basic argument is that how and when political parties are first organized has a strong effect on the role of political patronage. Externally mobilized parties will tend not to use political patronage. Internally mobilized parties, in contrast, will tend to use political patronage, unless they absolutist or progressive coalitions become entrenched prior to the mobilization of the masses into politics.
As such, Shefter makes a strong path dependent argument similar to Lipset and Rokkan (1967): social conditions at a key historical junctures can produce persistent effects on political institutions. Shefter's argument, alas, also demonstrates some of the weaknesses of path dependency. First, arguing that the entrenchment of absolutist and progressive coalitions played a determinant role in future political patronage systems only begs the question: why did some nations see the entrenchment of absolutist and progressive coalitions in the first place? Indeed, is the entrenchment of certain coalitions only an intervening variable on a much longer causal chain? Second, Shefter provides no argument for why the entrenchment of absolutist of progressive coalitions played a stronger causal role than say national culture, geography, or electoral systems. In short it is not enough to explain why a factor was important; it should also explain why a factor was more powerful than plausible alternative explanations. Third, while agreeing that political culture tends to be persistent, it is certainly implausible to argue that composition of a society in the 19th century should continue to predetermine political institutions forever. Clearly, political culture in societies change and a robust theory should be able to account for such change. Like many institutionalist scholar, Shefter's theory provides a better account of continuity than change. Lastly, Shefter presents weak empirical support for his argument. Of Shefter's five case studies (Germany, Great Britain, Italy, France, and the US), two (France and the US) present strong anomalies.
Acemoglu, Johnson and Robinson take an historical approach to explain path dependency of former colonies. They argue that the settler mortality rate determined the “institutional investment” of the colonial power. The results are quite interesting but the causality is not clear to me as they take a very narrow “historical” approach (one specific event in the past –settler mortality- and then discuss it within the current situation). The assumption that the institutional context was constant during the entire colonization period does not seem right especially when major external factors such as World War 1, the Great Depression have constrained investments in the colonies (in labor and finance) forcing the colonial powers to developed indirect ruling forms (mentioned by Greif and Laitin).
Greif and Laitin describe extensively the importance of external factors (economic and political pressure) in the case of Venice and Genoa on institutions at the local level.
There is one major point I hope we could discuss: The data. These readings refer (with more or less emphasis) to historical factors using data and information developed mostly by European countries (colonial powers in some cases) and I am just uncertain about the accuracy of “our” understanding of these sources. The different papers refer to norms that may have very different value in other contexts: is the notion of property rights considered as a universal? What is considered as basic criteria for these comparisons seem already oriented.
In Presidentialism, Parliamentarism and Democracy, Jose Cheibub attempts to refute the Linzian claim that the intrinsic features of presidential systems cause them to have less longevity than parliamentary systems. Instead, Cheibub argues that it is not presidentialism itself that is determining; rather, presidential systems are more fragile because they typically emerged in nations with a strong military presence, which “placed democracy of any kind at risk.” Do we find Cheibub’s claim convincing? Indeed, Cheibub offers no sound empirical test of his claim that “initial institutions …[are adopted] for reasons hat are unrelated to the ones that lead to the occurrence of military dictatorships” (147). Cheibub’s analysis is missing several causal links. While Cheibub describes the military/presidential nexus as an “historical accident,” he fails to explain what may influence a country’s propensity for militarization. Without that analysis, how can he claims that the reasons for militarism are not the same as for institutional development? Further, what causes a militaristic dictatorship to form rather than a civil dictatorship? We can also ask whether Cheibub’s work suffers from the same type of sample selection problem that Stokes and Boix point out in Przeworski and Limongi’s work. As noted in Chapter 4, Cheibub’s dataset is confined to the period 1946-2002. If we took the dataset farther back in time would the results change? Further, how sound is Cheibub’s theory about the importance of the Cold War in fostering the kind of militarism that leads to the breakdown of democratic regimes when most of his dataset consists of the time period of the Cold War? Would his analysis be different if he studied different waves of the Cold War, such as early Cold War, height of Cold War and post Cold War? Indeed, could we make the argument that the international environment at the time of regime formation rather than the presence or absence of a military regime may be more important in effecting the stability of presidential regimes? By confining his analysis primarily to Cold War cases, Cheibub is not differentiating between the importance of each variable. Finally, what can we learn from the Geddes article that may influence our understanding of why military regimes may cause subsequent democracies to be more unstable? Geddes argues that most military regimes end in negotiation as there are many internal disagreements and splits. It could be the case that this “symptom” of militarism rather than militarism itself plays a more important causal role in Cheibub’s analysis.
I argue that Martin Shefter's detailed historical analysis suffers from its small sample size. Selection bias is always a potential problem in these kinds of studies, but in this instance I think that the small sample size has allowed Shefter to obfuscate with respect to several ways of explaining variation in party and patronage across countries.
ReplyDeleteFirstly, he steadfastly refutes the neoclassical theory of patronage at the outset of his paper. He does this by citing several important counter-examples. However, at the end of the essay, when acknowledging the existence of important counter-examples to his own argument, he argues that his theory need not ‘utterly predetermine the subsequent character of a nation’s party politics’. Given this, is it wise to be so certain that the more ‘sociological’ explanations offered by the neoclassical theory do not also play a role?
Secondly, he offers little room for political agency in his account. Political actors who may benefit from systems of patronage may seek to find ways to supply offices and rents to supporters while maintaining the veneer of an independent bureaucracy or placating such (liberal or absolutist) forces as oppose patronage. The layers of special offices and positions created by the United States President with discretionary appointments are examples of this. In Ireland, similarly, an independent civil service existed at the time of independence in the early 20th century. The civil-service, largely based upon the British model, was retained in form and organization by successive Irish governments. This did not stop, however, successive governments from building layers of special commissions, advisors, officers, councils and the like on top of and alongside the regular civil service. These former positions provided the means for political patronage. Shefter’s discussion does not, I think, leave sufficient room for such hybridized cases.
Acemoglu, Johnson and Robinson’s use of mortality rate as an instrument for current institutions leads to interesting conclusions regarding the choices of colonial institutions and the persistence of those institutions throughout history. According to their conclusion, the influence of mortality rates—determined by condition in colonies—on colonizers’ decision to settle or exploit, and thus their choice of institutions to implement in the colonies, still exerts effect on economic development. This finding supports the belief that political institutions—particularly those of property rights—cause economic growth, and not the reverse.
ReplyDeleteFor Acemoglu et al., institutions persist. In fact, the lasting effect of extractive versus neo-European institutions has persisted through independence, across borders, and through various stages of state formation—be those states under democratic or authoritarian rule, post-independence. Their argument that institutions persist is not convincing, particularly when considering that many states included in their observations underwent institutional changes throughout their history. In particular, when contrasting this argument to their “Reversal of Fortune” article, published a year after “Colonial Origins,” it is exactly the shift in institutions that causes the reversal of fortune. If institutions persisted, then it would be difficult to convey an “institutional reversal.” However, colonial intervention created a disruptive shift, so that societies benefitting from higher levels of income and urbanization were left with extractive institutions, while in previously poor countries, institutions were created to foster investment. If we can consider the possibility that colonization can exert a strong shift, we can also consider the possibility that independence or a change in regime can exert a similar effect on institutions, so that institutions may not necessarily persist throughout these experiences.
Furthermore, the use of mortality rates as an instrumental variable may not accurately capture colonizers’ choices. Referring to Appendix Table 2, we see that Panama had a high settler mortality rate. However, as opposed to enacting extractive institutions, Spanish settled in the area, using Panama as a control base for neighboring colonies. On the other hand, South Africa, which displayed settler mortality rates close to that of the U.S. (15.5 versus the U.S.’s 15 and Canada’s 16.1), was colonized with extractive institutions, particularly after the discovery of diamond and gold in the region. Engerman and Sokoloff’s argument for the relevance of factor endowments in determining colonial institutions seems to be more appropriate, under this scenario.
One of the issues that I found especially interesting in this week’s readings is how to effectively measure institutions. Glaeser et al. criticize the three sets of measurements that have been commonly used in existing literature as “measuring only outcomes… not reflecting either constrains on government or permanent features of the political landscape”. Instead, they propose a more “deep” and precise measurement using electoral rules, judicial independence and constitutional review that are claimed to reflect permanent characteristics of institutions. Both measurements, the existing ones and the newly proposed one, are flawed in my opinion. Although the former ones that are constructed by survey data on perception of institutional quality and government effectiveness are comprehensive, they are inevitably contingent on interviewees’ cognitive ability and subjective bias; likewise, the way of selecting some indicators to proximate institutions is problematic because the selection per se is subjective and impossible to exhaust all “permanent” characteristics of institutions. The measurement problems of institutions can be attributable to the ambiguous definitions of institution and its extension. Take the dictator’s choices mentioned in Glaeser et al’s essay for example, should the dictators’ self-disciplining in expropriation and property rights violation as well as their developmental rationale be counted as a benign institutional constrain for economic development in some countries? The answer really depends on how we define “informal institutional constrain”. All of the institution measurements mentioned in Glaeser et al’s article are deficient in capturing some informal yet substantial institutional constrains in the reality.
ReplyDeleteThis comment has been removed by the author.
ReplyDeleteIn my mind, Acemoglu, Robinson, and Johnson’s (ARJ, 2001) original finding and Glaeser et al.’s (2004) alternative are both incomplete arguments, because argue deep institutional explanations based on the constraining power of initial conditions are key to development outcomes without offering even a cursory reason as to why the externalities imposed by such initial conditions should continue to hold over time. In effect, they use Hall’s (2003) initial path-dependent conception- in that initial conditions influence initial institutional outcomes that persist- without doing enough process tracing to show why institutions get “stuck”. It is too bad that AJR self-consciously leave institutions and the mechanism by which they persist as a black box, but Glaeser et al. spend a bit of time discussing the externality feedbacks generated by human capital, on the one hand, and institutions on the other. Given this, it is too bad that they don’t expand this a bit to show how initial human capital endowments shape incentives for favoring particular types of institutions that generate a feedback loop and allows both good institutions and high human capital to persist over time.
ReplyDeleteFurther study of both of these theories is a great opportunity for future work, because although institutions are widely regarded as sticky in the literature, an understanding of why and how this process occurs has been lost under broad generalities about “political culture”, “social mores”, and underspecified externalities. The sorts of deep institutional questions noted above allow for unique opportunities to disaggregate parametric requirements (a la Grief and Laitin, 2004) for equilibrea in order to get a better grasp of which factors tend to be more central to strategic choices at particular times by allowing for long observations. For example, in the case of Glaeser et al., where the key determinant of such parameters is posited to be human capital, it should be possible to craft a research plan centered around understanding how changes in human capital endowments may have shifted incentives for particular types of institutions and, more interestingly, why such shifts did not occur uniformly with increasing human capital endowments.
The discussion over the relationship between economic growth and institutions advanced by Acemoglu, Johnson and Robinson (AJR), as well as Glaeser et al. is a perfect example of the frustrating nature of the "chicken or egg" debates that dominate this issue. Both papers provide compelling empirical evidence for their findings, but their conclusions are contradictory. By pointing out, for example, glaring problems with the traditional measurements of institutions (i.e., confounding institutions with their outcomes), Glaeser et al. show that, even if AJR’s methodology is sound (which itself is questionable), the chosen variables are biased. This conversation demonstrates the ambiguity and difficulties associated with empirical analysis of social phenomena. That said, if scholars choose this particular type of variable-based methodology, it seems that the discourse will simply descend into a never-ending commentary on such important, yet inconclusive and tangential issues as variable selection and measurement bias.
ReplyDeleteHall provides a refreshing perspective on this dilemma, as well as a convincing strategy for resolving it. Since interactions and dynamic processes make social phenomena too complex for traditional regression analysis, political science scholars should focus on devising a methodology that "catches up" to the current ontology that incorporates strategic decision-making and time elements. It seems persuasive to me that focusing on detailed, small-N studies can yield valuable information about processes that would enhance discussions such as that between AJR and Glaeser, et al. In fact, as other scientists already recognize, detailed examinations of particular cases are bound to reveal consequential causal information that would otherwise be assumed away or overlooked in traditional macro-level regression analysis. Furthermore, it would wean political science scholars off the tendency to ascribe more causal weight to statistical inference than traditional regression models can bear.
Glaeser et al. argue that three of the measures of institutions from the economic growth literature do not, in fact, measure institutions. Their argument creates doubts about the proposal that economic development can be attributed to institutions. One of the measures they highlight is the risk of expropriation used by Acemoglu, Johnson, and Robinson. Glaeser et al. argue that this measure, along with a measure of government effectiveness, measures outcomes instead of institutions. Their definition of institution requires that they are permanent or durable constraints on government. AJR's measure is neither a constraint (as it includes the free choices of dictators) nor permanent (it rises with income and is highly volatile). Glaeser et al. propose an alternative measure of institution, one that does not look at assessments of institutional outcomes, but rather manipulable rules and procedures. While their suggestion would move the measure closer to their definition of institution, it does not fully explain why institutions are persistent---or why they can change, giving space for an incorporation with Laitin and Grief's analysis.
ReplyDeleteI appreciate very much Greif and Laitin's purposes in presenting their 'Theory of Endogenous Institutional Change'--both the overall goal of joining the strengths of historical institutionalist analysis and game theory for a more comprehensive approach to institutions, and the more precise goals of creating a framework in which (1)endogenously created 'variables' can act like (exogenous) 'parameters,' and (2) institutions reinforce themselves by widening the range of situations in which the behaviors they engender will act like exogenous parameters to reinforce the institution on its current path. This framework is helpful because it allows an institution to be successful at 'enforcing' the behaviors they entail but be unsuccessful at reinforcing itself. But there are a few points I would like to discuss and have clarification on from any of you who understand this piece well:
ReplyDelete1) What, clearly, is the difference between self-enforcement (is this just enforcing the behaviors they dictate/entail?) and self-reinforcement?
2) Do Greif and Laitin 'strawman' historical institutionalist uses of critical junctures? Misuse of this construct seems to be one of their key complaints with historical institutionalism.
3) What does this framework really add to a potential institutional analysis by putting the framework in game theoretical terms? The framework may certainly correct some key mistakes in the historical approach, but much of it seems doable without game theory explicitly. Is this (1) a new and significant contribution, or (2) more simply an attempt to make relevant institutional approaches accessible to the game theory - oriented community in political science, and/or (3) an attempt to make game theory approaches to institutional analysis the dominant ones? (As a bridging exercise between the two broad approaches, it is helpful. But I'm left wondering whether it does much that a clearer, more spelled-out historical institutionalist model could not.) Any ideas?
Leanne Tyler Comparing Institutions Week Three—Institutions, Endogeneity, and Path Dependency: Comment
ReplyDeleteBoth Acemoglu et al. (“The Colonial Origins of Comparative Development) and Glaeser et al. (“Do Institutions Cause Growth”) attempt to use path dependent arguments to account for divergent levels of economic growth in the developing world. The implication of their research project is that historical developments dating back to the colonial period have had a longstanding impact on current levels of development in some of the world’s poorest countries. Yet, both of their articles seemed to “miss the mark” in terms of providing a compelling explanation of how their independent variable of interest (“institutions” and “human capital”) are causally related to economic growth. In the case of these two pieces, it seems as though another research method, perhaps one involving a more robust discussion of historical developments through process tracing would have been more appropriate. Both of the authors’ insistence on validating their argument using statistical tests resulted in the neglect of any discussion of the complex political and economic processes underlying the correlation that they identify in their respective studies. These approaches are not mutually exclusive. I would argue that the use of both would have made the arguments of the authors much more compelling. Furthermore, it would seem impossible to gain a better understanding of countries that do not fit into their theoretical models without utilizing such an approach. This tendency to propose historical arguments with few or no case studies or comparative analysis is a good example of the disjoint that Hall discusses between ontology, which acknowledges the importance of historical causal factors, and methodology, which is increasingly favoring statistical analysis, in political science.
This weeks readings substantively discuss issues of institutional change – the how, why, when, consequences of, and rationale for shifts in institutional structure. More fundamentally however I was interested in the manner in which a subset of readings deal with issues of research methodology in the political sciences – by either questioning the validity of traditional constructs in statistical methods (Glaeser et al. 2004; Acemoglu et al 2001); exploring the need for alternative approaches that balance rational and structural methods (Kitschelt 2002; Hall 2003); or improving and refining existing concepts of path dependence (Page 2006; Laitin and Grief 2004). As a newcomer to political science it was interesting to see the debate on institutions remain so centered on issues of process, and the degree of ambiguity that remains. How far has this debate evolved?
ReplyDeleteGreif and Laitin provide a framework for conceptualizing endogenous institutional change. The concept of quasi-parameter (parameters that over time are changed by institutions and in the long run effect the choices of actors) allows us to think of how institutions can have different short and long term consequences depending on their effects on on these quasi-parameters. It would be interesting to apply their frame work the the Latin American colonial experience described by Engerman and Sokoloff. Although under the rule of the same overall spanish colonial institutions characterized by extreme inequalities, differences in the how authority or grants were distributed among settlers between the different regions could have led to different long term outcome. While those institutions were self-enforcing in that they resulted in economic gains that ensured elites would have an interest in the perpetuating them, the above mentioned differences could have compounded with the growing wealth of elites and led to different outcomes were external parameters to change, for example. Competition among elite groups could lead to the coopting of 'the people' in some scenarios as spanish rule declined, could have undermined the self-reinforcement of the unequal institutions. In general I found Greif and Laitin's theory very useful as it is able to incorporate into an overall game-theoretic framework the very realistic assumption about how institutions, rules, themselves can alter the beliefs and payoffs of different actors over time.
ReplyDeleteAcemoglu et al. (2001) and Glaeser et al. (2004) posit alternative theories concerning the importance of institutions in explaining economic growth. Each authors’ attempt to offer a “deep” explanation of causal change falls short. Acemoglu et al. fail to provide an historically accurate context for their argument. In the first instance, they try to compare all ex-colonies as if they were settled in the same way in the same time and place. The settlement of Latin America in the 1500s differed dramatically from the way Africa was settled in the 1800s. The world had changed; the Industrial Revolution brought with it new technologies and advances in medicine impacted diseases. Further, institutions in the colonizing nations had changed. Institutions that protected private property were only beginning to surface in Europe during the initial waves of colonization and continued to be perfected over time. Acemoglu et al. also do not specify the criteria they use to differentiate the two types of colonizing types. Indeed, might there be more a graduation in approaches, such as degrees to which institutions were “extracting”? Acemoglu et al. also fail to fully specify how their path dependence argument works – are they making an argument about the importance of path dependency or phat dependency?
ReplyDeleteLikewise, while Glaeser et al. provide a compelling argument, they too underspecify how their theory operates in practice. Indeed, they admit that they do not want to “push [their] results too far,” (293) and that a myriad of other factors could have impacted the colonization process. How is Glaeser et al. a robust theory then if, as they freely admit, their potential instrument term (human capital) may be correlated with the error term? Further, how should the term “human capital” be defined? Glaeser et al. define in primarily in terms of years of schooling. Might there be other variables, such as literacy rates or variables based on skills that may be a better instrument to study in which to measure “human capital”? Glaeser et al. also question Acemoglu et al’s method of institutional measurement. But Glaeser et al.’s measures too suffer from collinearity issues. Further, their proposed measurement (electoral rules, judicial independence, constitutional review) are not as stable in many of the colonized nations as Glaeser et al. assume. Amy Semet
In terms of institutional change and diversity, Herbert (2002) points out the importance of an analysis including both structure and human agency. Avner and Laitin (2004) construct a model which introduces ideas from the game theory and historical institutionalism. Hall (2003) proposes a “systematic process analysis” to cope with the gap between theoretical ontology and methodology. All the above works intend to combine the pros and address the cons of the current research approaches. Yet, none of them provide a satisfactory theory/approach to study institutional change. Herbert simply concludes with the impossibility of prediction. I’m not sure if I’ve learned how to differentiate “variables,” “parameters,” and “quasi-parameters.” Institutions are endogenous. AJR (2001) and Glaeser et al. even show reverse directions of causality regarding the relations between growth and institutions. Since I am incapable of offering a better solution about the study of institutions, I decide to say something about Shefter’s article. It is a historical analysis of group politics. It seems to me that the analysis has less to do with the core assumption of “internally and externally mobilized parties” and their preferences for and against patronage. Besides, with a focus on early group coalition, I think the author neglects the effect of electoral rules on the party’s strategy to grant patronage.
ReplyDeleteThe literature on the effects of institutional design on economic growth has demonstrated that “institutions matter,” but the exact mechanisms by which institutions affect development across countries remain largely unspecified. In AJR (2001), too, the question of which and how institutional processes affected economic (under)development remains unanswered. The causal argument in their paper merely asserts that colonizers established institutions conducive to growth where environmental conditions rendered settlement possible, and that they established extractive institutions otherwise.
ReplyDeleteBut what about cases where colonizers might have encountered both conditions conducive to settlement and high levels of resources? AJR’s settlers would have set up institutions that protected property rights and conferred the benefits of resource extraction to the ruling elite. The initial allocation of economic rents to a small group would have discouraged institutional developments conducive to growth and development, and thus despite property rights, limited access to economic opportunities for the majority of the population would have hindered economic development (Engerman and Sokoloff, 2002). Distributive factors independently affected institutional outcomes and their effects in turn, even when settler migration occurred.
Hence, what AJR summarize in a single causal argument is in fact much more complex. Institutions are shaped and change as a result of geographic, economic, political, and social conditions; analyses of the causal factors of economic development must focus on the micro-processes within countries that take these complex conditions into account. In this context, the effectiveness of broadly focused cross-country regression analyses is limited given that instrumental variables such as settler mortality rates for institutions are often too coarse to permit a close analysis of the causal mechanisms linking institutions and economic development; the invariability of instruments like settler mortality furthermore limits their effectiveness in explaining institutional change.
According to Kitschelt, a good social science explanation has two features: it provides “deep” causes, and at the same time offers microlevel mechanisms that account for the relationship between the independent variable(s) and outcomes. Both features are a call for taking issues of temporality seriously. What comes first and last in determining outcomes (the order of events, or its sequence) is crucial in theory construction in general, and especially important in theories dealing with (endogenous) institutions. How well do explanations dealing with institutions score on both these fronts? Acemoglu et al.’s story goes deep (colonial history) but does not perform well at the level of mechanisms (e.g. how is that old institutions affect new ones?) In contrast, Glaeser et al. explanation is shallow (human capital is too “proximate” a cause of development) and the mechanisms unclear (why do some autocracies choose to invest in education and others not?).
ReplyDeleteIn "Party and Patronage", Martin Shefter seeks to account for international variations in the role played by political patronage. Shefter's basic argument is that how and when political parties are first organized has a strong effect on the role of political patronage. Externally mobilized parties will tend not to use political patronage. Internally mobilized parties, in contrast, will tend to use political patronage, unless they absolutist or progressive coalitions become entrenched prior to the mobilization of the masses into politics.
ReplyDeleteAs such, Shefter makes a strong path dependent argument similar to Lipset and Rokkan (1967): social conditions at a key historical junctures can produce persistent effects on political institutions. Shefter's argument, alas, also demonstrates some of the weaknesses of path dependency. First, arguing that the entrenchment of absolutist and progressive coalitions played a determinant role in future political patronage systems only begs the question: why did some nations see the entrenchment of absolutist and progressive coalitions in the first place? Indeed, is the entrenchment of certain coalitions only an intervening variable on a much longer causal chain? Second, Shefter provides no argument for why the entrenchment of absolutist of progressive coalitions played a stronger causal role than say national culture, geography, or electoral systems. In short it is not enough to explain why a factor was important; it should also explain why a factor was more powerful than plausible alternative explanations. Third, while agreeing that political culture tends to be persistent, it is certainly implausible to argue that composition of a society in the 19th century should continue to predetermine political institutions forever. Clearly, political culture in societies change and a robust theory should be able to account for such change. Like many institutionalist scholar, Shefter's theory provides a better account of continuity than change. Lastly, Shefter presents weak empirical support for his argument. Of Shefter's five case studies (Germany, Great Britain, Italy, France, and the US), two (France and the US) present strong anomalies.
Simon Collard-Wexler
Julie Touber
ReplyDelete(Originally posted February 8, 2009 9:30 PM)
Acemoglu, Johnson and Robinson take an historical approach to explain path dependency of former colonies. They argue that the settler mortality rate determined the “institutional investment” of the colonial power. The results are quite interesting but the causality is not clear to me as they take a very narrow “historical” approach (one specific event in the past –settler mortality- and then discuss it within the current situation). The assumption that the institutional context was constant during the entire colonization period does not seem right especially when major external factors such as World War 1, the Great Depression have constrained investments in the colonies (in labor and finance) forcing the colonial powers to developed indirect ruling forms (mentioned by Greif and Laitin).
Greif and Laitin describe extensively the importance of external factors (economic and political pressure) in the case of Venice and Genoa on institutions at the local level.
There is one major point I hope we could discuss: The data. These readings refer (with more or less emphasis) to historical factors using data and information developed mostly by European countries (colonial powers in some cases) and I am just uncertain about the accuracy of “our” understanding of these sources. The different papers refer to norms that may have very different value in other contexts: is the notion of property rights considered as a universal? What is considered as basic criteria for these comparisons seem already oriented.
In Presidentialism, Parliamentarism and Democracy, Jose Cheibub attempts to refute the Linzian claim that the intrinsic features of presidential systems cause them to have less longevity than parliamentary systems. Instead, Cheibub argues that it is not presidentialism itself that is determining; rather, presidential systems are more fragile because they typically emerged in nations with a strong military presence, which “placed democracy of any kind at risk.” Do we find Cheibub’s claim convincing? Indeed, Cheibub offers no sound empirical test of his claim that “initial institutions …[are adopted] for reasons hat are unrelated to the ones that lead to the occurrence of military dictatorships” (147). Cheibub’s analysis is missing several causal links. While Cheibub describes the military/presidential nexus as an “historical accident,” he fails to explain what may influence a country’s propensity for militarization. Without that analysis, how can he claims that the reasons for militarism are not the same as for institutional development? Further, what causes a militaristic dictatorship to form rather than a civil dictatorship? We can also ask whether Cheibub’s work suffers from the same type of sample selection problem that Stokes and Boix point out in Przeworski and Limongi’s work. As noted in Chapter 4, Cheibub’s dataset is confined to the period 1946-2002. If we took the dataset farther back in time would the results change? Further, how sound is Cheibub’s theory about the importance of the Cold War in fostering the kind of militarism that leads to the breakdown of democratic regimes when most of his dataset consists of the time period of the Cold War? Would his analysis be different if he studied different waves of the Cold War, such as early Cold War, height of Cold War and post Cold War? Indeed, could we make the argument that the international environment at the time of regime formation rather than the presence or absence of a military regime may be more important in effecting the stability of presidential regimes? By confining his analysis primarily to Cold War cases, Cheibub is not differentiating between the importance of each variable. Finally, what can we learn from the Geddes article that may influence our understanding of why military regimes may cause subsequent democracies to be more unstable? Geddes argues that most military regimes end in negotiation as there are many internal disagreements and splits. It could be the case that this “symptom” of militarism rather than militarism itself plays a more important causal role in Cheibub’s analysis.
ReplyDeleteAmy Semet