Shleifer and Vishny argue that corruption can take place under three environments: joint or monopolistic environment, a multiple independent monopolistic environment, and a competitive monopolistic environment. They go on to predict that the level of revenue collected in each (and the attendant distortion caused by the corruption) in respective descending order of magnitude is: the second case, the first case, and the last case.
One way to test the theory is to leverage differences in the structure of regulatory responsibility across related, but distinct, regulatory areas and explore how corruption influences the ability of business people to operate under each. An excellent example can be found in a related work by Martin Dimitrov on intellectual property (IP) rights in China. Dimitrov classifies the regulatory scheme for patent, trademark, and copy right law as well as the ministries involved in enforcing each in China. Although he is interested in rule of law, he shows that each of these branches of IP law have differing numbers of regulatory agencies in charge of them, complicating enforcement in the case of trademark and copyright law where multiple agencies have responsiblity but simplifying it for patent law, where only one agency is involved. Using this as a base, one could then survey businesspeople in the related area in order to measure the level of distortion caused by corruption for each type of IP law (or whatever regulatory area one choose). Alternatively one could leverage cross-regional variation in local regulatory models for a single issue, such as building or business permits. This has the advantage of posing more controls, although it might be harder to conduct because one would need to classify regulatory procedures in a large number of local settings.
In either case, the measure of corruption would be the number of different agencies or groups one needs to speak with in order to obtain the regulatory good sought. This could then be used as an independent variable, along with controls for level of development, partisanship (in countries where it applies), etc. in a regression with the measure of corruption obtained from looking at the cross-regulatory (or cross regional) variation in the number of actors involved in obtaining the regulatory good.
My apologies, I accidentally posted before completing the last paragraph. It should read: In either case, the measure of corruption would be the number of different agencies or groups one needs to speak with in order to obtain the regulatory good sought. This could then be used as an independent variable, along with controls for level of development, partisanship (in countries where it applies), etc. in a regression with the measure of distortion obtained through surveys of business people who are involved in that particular regulatory area. Possible instruments might include asking how much respondents believe corruption hampers their business, asking them to estimate the costs of such corruption, etc. (questions similar to this have been asked in the European Bank on Reconstruction and Development and World Bank's Business Environment and Enterprise Performance Survey in Russia). If this is the dependent variable, then one would also want to control for enterprise level features such as firm size, etc.
References (forgot to post them)
Dmitrov, Martin (2008). Piracy and the State: The Politics of Intellectual Property in China. Cambridge: Cambridge University Press.
Dmitrov, Martin (2004). Ph. D. Dissertation: Administrative Decentralization, Legal Fragmentation, and the Rule of Law: The Enforcement of Intellectual Property Rights Laws in China, Russia, Taiwan, and the Czech Republic. Stanford University.
Survival (duration) analysis would be one way to empirically test Geddes’ hypotheses on the effects of political competition on civil service reform. In this analysis, the dependent variable would be the timing of policy adoption. The individuals in the sample are countries, which are observed when they enter the sample and up to their exit from the sample or failure (here, countries fail when they decide to adopt civil service reforms). Among the independent variables, one would include factors such as: a) % of seats heald by the main parties in the legislature, or better still, b) the difference in seats between the incumbent’s party and the largest opposition party in the legislature, as a proportion of the total number of seats (see Murillo and Gallardo AJPS 2007 for such an empirical strategy). These measures would capture the relative power of parties in the policymaking process which is the key mechanism affecting the incentives of politicians to reform in Geddes’s model. Finally, a full set of controls are needed: these are not only political in nature (such as the ideological distance between parties, or the effective number of parties) but also economic (whether countries face sever fiscal constraints, and the rate of economic growth). Ideally, the sample would include all Latin American countries (not just 5 cases) and the time period could we extended to include the most recent democratization wave (this would fulfill the Geddes condition of studying 15 or more years of democratic stability). As a complement it would help the model a lot to find systematic relationships between the independent variables discussed above and blunt patronage measures such as the share of public employment in the economy or the contribution of salaries in total public expenditures. We could imagine Graphs that show how changes in the difference in seats between the largest parties are correlated or not with changes in patronage outcomes.
Geddes observes that the access to patronage is closely related to electoral system (i.e. PR system with an open or closed list). Reforms are possible and sustainable only when the top (two) parties possess similar legislative clout (access to patronage) and high party discipline. She concludes that administrative reform is difficult in fragmented party system. Both the party system and party discipline are endogenous to electoral rules. A follow-up question will be: under what conditions does the above equilibrium happen? When do the politicians need to depend on their personal votes instead of party votes? One condition referred in the Geddes piece is the PR system with an open list. Another condition I can think of is the multi-member-district under majoritarian rule, which often leads to intra-party competition. One implication from Geddes’ paper is that in a bi-party system with strong party discipline, administrative reform is most likely to be successful. Can anyone think of a counter-example?
Marcus Kurtz and Andrew Schrank (KS) revisit the causality between institutions and economic growth. They see economic growth as the cause of good governance and not the opposite. In a response to KS, Daniel Kaufmann, Aart Kraay and Massimo Mastruzzi (KKM) defend the opposite: good institutions are the cause of sustained economic growth.
Identifying the relevant variables to evaluate good governance are debatable, and KS and KKM also engage in a debate on the perception of economic development, that according to KS is too limited to the business community. Timeline is also a key variable that is debated between KS and KKM. KKM point out the importance of long-term economic growth and reproach KS for looking at a very short period of time. It seems to me that KKM arguments are more plausible than KS’s one for this very last reason. However, the weakness of KKM argument is that it does not provide for an explanation of how some countries came to have good governance and from a policy point of view is quite pessimistic as it implies that no evolution is plausible.
One way to contribute to this debate and test parts of these arguments could be to look at foreign aid as “economic growth” and see which countries can sustain this growth or not, and for what reasons. I feel that absorptive capacity could be considered as a proxy for government effectiveness.
Cooter (2000) develops a delegation game to determine how a rational principal makes the decision to either directly exercise power or delegate it to an agent. The principal’s decision hinges on the probability of a good state of nature. Good luck can disguise diversion, while bad luck reveals it. As the probability of a good state of nature increases, the principal will be less likely to delegate. We could test Cooter’s prediction by examining the random event, the state of nature.
In the game, the state of nature cannot be controlled or predicted by either the principal or the agent: “To illustrate, most state administrators cannot predict or control elections or the stock market” (Cooter 2000, 81). We could examine events that, in retrospect, would have changed the probability of a good state of nature (crisis, election, etc.) to determine if the principals in fact did change their level of delegation in response. We would expect that a shift to a good state of nature would decrease the level of delegation, while a shift in the opposite direction would increase the level of delegation. Ideally, we could use one probability-changing event across bureaucracies to control for the possibility that the nature of the event itself affects the principals' behavior. We would also have to control for the type of bureaucracy (if some are more sensitive to changes than others).
In Deliberate Discretion, John Huber and Charles Shipan attempt to develop a theory of how laws shape bureaucratic autonomy. They use legislative word count as their dependent variable. For instance in their example concerning the Medicaid program, they use as the dependent variable the change in the number of words in the statute in 1995-1996. There are a number of problems with this measure. First, word count may not necessarily be a good proxy for discretion. Looking at a statute in isolation may not give all the details concerning how discretion is delegated. A statute may refer to or incorporate another statute or existing regulations. Statutes in different policy areas or in different languages may just simply use differing amount of words. Legislators may also have different proclivities toward the number of words they have in a specific legislative proposal. These problems are attenuated when one uses changes in word count as a proxy for discretion. Huber et al’s measures does not accurately account for whether changes in the statute are due to varying levels of discretion or if they are due to some other variable. Although Huber et al. recognize this problem and attempt to control for it somewhat, there are still remaining problems with this measure. For instance, if a statute changes by 1000 words how does one differentiate between situations where the 1000 words were added or if the original statute contained 2000 words and 1000 words were deleted? Huber et al’s measure also self-admittedly acknowledges that use of word count as a proxy is very circumstance-dependent.
How then can we create a better measure of discretion in order to test how laws shape bureaucratic discretion? This is a difficult problem and the alternative measures proposed suffer below from many of the same infirmities as the Huber et al. measure. Nevertheless, these alternatives are at least necessary to consider. Instead of focusing on statutes, one could focus on the scope, number and content of regulations. This, of course, relies on the assumption that administrative agencies have a custom of adopting regulations, as they do here in the United States. If this assumption is not true, then, this model could perhaps only be useful to measures discretion in places like the U.S., where regulations are freely promulgated. One could assess whether a greater number of regulations means that the agency has more discretion to promulgate its own rules with respect to the issue at hand. Although this too is a very-issue specific measure, it could perhaps be a better proxy for discretion. OLS or even negative binomial regression could be used to estimate the results. One could also measure bureaucratic discretion by creating an index. For instance, one could individually measure whether the bureaucracy has the ability to do certain things, i.e., they have the authority to promulgate regulations and how many of these regulations have been overturned, whether they have the authority to create commissions, etc. This index could also incorporate restrictions on the agency, such as whether the agency is subject to judicial review, sunset provisions, legislative oversight, etc. The codings on the various issues could then be summed up and each agency could be given a “score” on how much discretion it is given the legislature, on a specific bill, or collectively on many bills. There are still other possibilities for how we could measure discretion. One could measure it by the number of employees as agency has, speculating the discretion increases with the number of employees. One could also measure discretion by the amount of the agency budget, with the assumption that the greater the budget the more discretion the agency has.
Another problem with Huber et al’s measure is that it places primary importance on the legislature as the source from which discretion for the agency emanates. However, many agencies owe their allegiance to the executive branch, and thus, focusing solely on legislative instruments such as legislative word counts would not fully account for how much discretion the agency has. Thus, moving away from a legislative measure to a measure that focuses more on the agency itself (such as one of the measures suggested above) may paint a more accurate portrait of discretion, at least for these types of executive agencies. Of course, measuring the effect of law on discretion in executive agencies would require the analyst to employ different independent variables than the ones Huber et al. imply (though some, such as divided/unified government. etc could still be employed)
The debates between Kurtz & Schrank (KS) and Kaufmann et al. (KKM) focus on how to effectively measure “good governance”. KS criticizes KKM’s popular measures of government effectiveness (GE) being problematic because 1) they conflate policy preferences and outcomes with political institutions and 2) relies almost entirely upon the perceptions of businesspeople whose views of governance might be different from other groups of people. I think that the flaws of KKM’s measures can mostly be attributed to its subjective nature. The weaknesses of the subjective GE measure would be hard to eliminate even thought KKM had plausibly taken a number of modifying measures such as weighting the data collected from different sources/groups of people and framing the wording of survey questions as to assess the likelihood of governments’ expropriation rather than the actual expropriation etc. The subjective assessment of governance which mainly constitutes KKM’s GE measure has to more or less base on respondents’ previous experience to deal with the government and their existing perception of governance before the surveys were conducted. It is to be noted that people’s experiences and perceptions of governance are inevitably influenced by the economic performance of the time. As a result, KKM’s measure of governance effectiveness relates to previous economic growth by construction while it has been demonstrated to be important for future economic growth by “a large and careful” economic literature. No wonder other scholars including KS question the causality order between the two variables: good governance and economic growth. Temporal order between two variables is one of basic evidences by which we ascertain causality order between them. Since the subjective nature of KKM’s measures obscure the temporary order between good governance and economic growth, they are not helpful for us to sort out the causality arrow between these variables. In this sense, it is tempting to use objective measures of governance. Then the question becomes: can we objectively measure governance? If yes, what would the criteria of good governance be?
To my knowledge, there are some quantitative and objective indicators measuring state capacity in existing literature and they might lend insight into measuring governance. For example, some scholars measure state capacity by the ratio of budgetary revenue/expenditure shares between the central and local governments (Wang and Hu, 2001). Because of the ratio form, the potential impact of the temporal economic performance on government’s budgetary or tax revenues can be controlled for. Because almost every government in the modern world has government finance statistics computed by relatively unified standard, such objective and quantitative measures of governance might be desirable in large-N cross-national comparison. Of course, whether objective measures like this can do a better job in measuring “a more basic notion of governance going back to the seminal work of Douglas North: the norms of limited government that protect private property from predation by the state” (KKM, 2007: 553) than KMM’s subjective one requires further validation.
Reference Kaufmann, Daniel; Kraay, Aart; Mastruzzi, Massimo. "Growth and Governance: A Reply." the Journal of Politics 69, no. 2 (2007): 555-62. Wang, Shaoguang; Hu, Angang. The Chinese Economy in Crisis: State Capacity and Tax Reform: East Gate Book, 2001.
Kurtz and Schrank argue that the understanding of the relationship between 'good governance' and economic growth has been made difficult by a general acceptance of the causal arrow going from the former to latter without conclusive evidence. What evidence there is, they claim, is potentially biased in that 'good governance' is defined from the point of view of businesses (whose interests may not align with that of citizens in general or of the state). They go on to argue instead that evidence seems to point to the inverse causal relationship: growth leads to better governance. One way to test this argument may be to begin at a national and sub national level (as the authors point out, perceptions of governance quality may be influenced by cultural perspectives, using these units of analyses would help eliminate this distortion). While it is difficult to come up with a universal and testable definition of what constitutes good governance, finding certain policies that would fall under such a category (though not be its full universe) shouldn't be that hard and may prove easier to test. The test that I am thinking of would then be to find a policy that can be defined as constituting a good governance practice that is articulated at the national level (this makes more sense in a federal system), and then to see whether implementation at the local/state levels precedes or follows any changes in economic growth at the subnational level. For example, beginning in the 90's in Brazil(following a national level articulation of the need to better State-citizen relations), state governments began creating government bodies that would unite in one location (per municipality) citizen access to information and services relating to the government (driver's licenses, government ids, applications for unemployment, pensions, criminal records, consumer complaints, etc.). The timing and extent of implementation and expansion of these bodies varied from state to state, with some beginning in 1991 while others as late as 2003. If economic growth positively affects good governance practices, can (preceding) growth at the local/state level account for the timing of the creation of these organs?
Shleifer and Vishny (1993) employ the idea of modelling corruption as the outcome of demand and supply forces within an unofficial market for government goods and/or services and apply insights from industrial organization. They proceed to argue that corruption can take place in three different kinds of markets: monopolistic, competitive, or one with independent monopolists. Their mechanism for determining the effects of corruption in these various contexts focuses on the degree of coordination, or lack thereof, between public officials, whom private firms need to bribe to obtain licenses (or permits etc.) that are complementary to each other. They distinguish between two types of corruption that these officials may engage in: corruption with theft and corruption without theft. (Though I will not go into it here, it might be an interesting way to extend Shleifer and Vishny’s theoretical insights by exploring how they hold up in situations where public officials deal with corruption opportunities that are essentially unrelated).
Shleifer and Vishny do not lend much empirical support to their claims, although they refer briefly to various historical cases. One way of testing their argument would be to look at levels of investment in countries with high levels of corruption vis-à-vis those with low levels. Based on their argument, we would expect to find that levels of investment – whether by the state itself, FDI, or by private actors within the country – would be lower in countries with high corruption, and thereby also find less technological innovation and perhaps even entrepreneurship, than in those with low levels. They cite Mauro’s (1993) finding that “holding 1980 real GDP constant, countries with higher corruption have a lower ratio of both total and private investment to GDP.” (p. 611). Further, they mention the problem and lack of foreign investment in post-Communist Russia and relate it to their model of a market with independent monopolists (p. 615). This example suggests the plausibility of their model. An updated and more rigorous empirical analysis, across a wide sample of countries, including data from the last 15 years (to the extent that this is available) would be a way of testing the model.
On a separate note, it might also be interesting to test Shleifer and Vishny’s arguments by examining entry and exit of firms in given sectors – construction, say, where government contracts, permits or licenses will need to be obtained by firms in order to operate, and could be subject to bribery or other forms of corruption – in countries with medium to high levels of corruption throughout the world (e.g. Nigeria, China etc.). Shleifer and Vishny themselves remark how a possible outcome of corruption with theft might be that firms who pay bribes will survive in any given industry or market and those who do not will go out of business (due to levels of costs, access to licenses, materials etc. essential to do business and so forth). We would therefore expect to see that firms that pay bribes in this context would be more likely to survive than firms that do not. Constructing an appropriate measure for this that captures the effect of corruption on firms’ fortunes might be difficult and it would certainly be a difficult task to establish causality (i.e. might other effects have an impact on the fortunes of the firm and not simply the impact of corruption? If so, which?). Still, it seems that this would be one fruitful way of empirically testing the implications of their argument, if an appropriate strategy can be devised.
Shleifer and Vishny’s propositions about corruption sit well with Cooter’s and Huber’s general arguments that relate institutional capacity to agency discretion and in turn corruption, which Shleiger and Vishny as well as Geddes argue effect economic development and growth adversely. As far as public sector services constitute input factors for private production, however, more efficient supply of government services through bribes might increase private sector productivity. An interesting question arises as to the extent to which the private sector surplus that arises from more efficient provision of public sector services might not only positively affect economic growth but also tolerate higher government taxation to offset losses due to corruption. In addition, to what extent might corruption be able to circumvent excessive regulation that itself was set up to increase corrupt rent-seeking by bureaucrats? More detailed models that study the effect of corruption across different levels of the bureaucracy might be able to detect when corruption by one agent (or in fact by the principal) at a higher level in the bureaucracy generates incentives for the emergence of corruption by other agents on lower levels of the bureaucracy? These models could advance our understanding of the effects of “corruption to evade corruption” on private sector output in developing and transitioning economies that feature complex corruption structures across several levels of the administration.
Shleifer and Vishny (1993) evaluate how government structure impacts the prevalence and nature of corruption. The conclusion of their analysis is that corruption can be reduced by political and economic competition – they argue that corruption is minimized where “several complementary government goods can be supplied by at least two government agencies” (606).
This argument could be tested by examining a specific sector such as the construction industry across selected subregions and/or countries. The construction sector is particularly suitable because the ‘lumpy’ nature of investments, as well as the many bureaucratic permitting process make it especially prone to corruption. It is also a sector where one could envisage ‘multiple windows’ for government services such as building licenses and permits. To test the theory, the independent variable of interest would be the ‘plurality of licensing/permit processes” – i.e. a measure to determine the ‘competitiveness’, or number of government agencies in providing building permit and licensing approvals. The dependent variable would then be a measure of sector-specific corruption. This could be constructed along similar lines to the World Bank Governance index – comprising a mix of indicators of the perceived and measured probity and efficacy of government agencies relevant to the construction sector (Kaufman, Kraay and Mastruzzi, 2007). It is not clear that data is readily available on these measures. It is likely that they would need to be collected through field visits and interviews. Once date is collected however a regression analysis could be conducted to determine the correlation between ‘plurality’ and levels of corruption.
Cooter’s argument about the “delegation chain” is interesting, yet unsubstantiated by the type of empirical evidence that could lend it validity. His claim is that, as you move up the chain of power in a state’s administration (from the low-level bureaucrat to the top-level elected official), discretion increases. According to his analysis, since top officials often face a volatile policy environment, the costs associated with formal constraints would be prohibitively high to ensure effective policymaking. Furthermore, the level of scrutiny of top-level officials, in addition to their desire for reelection, reduces the incentive to be corrupt. For low-level bureaucrats, inflexibility is more appropriate since they face relatively stable policy environments, and are unmonitored by and unaccountable to the general public.
It would be interesting to test this hypothesis in light of what we know about the technical expertise of lower-level bureaucrats. For policy issues where implementation is highly contingent on this technical expertise, would we really expect large constraints on low-level experts? In fact, on some highly technical policy matters, it seems that the low-level bureaucrats might actually have a higher degree of knowledge than the Secretary, or Minister of the relevant Department or Ministry. This would imply discretion increases as you go down the chain.
One approach to test this hypothesis would be to utilize the statutory analysis framework designed by Huber and Shipan in Deliberate Discretion? Because we are interested in looking at whether there is a higher level of discretion given to top-level officials than lower-level officials, we could examine regulations issued by departments, rather than statutes passed by a legislature. One promising area might be securities regulation, since independent regulatory bodies often govern it and it involves a continual and complex flow of technical information. We might expect that low-level bureaucrats would have a large degree of latitude in determining, for example, whether to approve new securities registrations. We could design an analysis of regulatory procedure across countries for new offerings by coding regulations by the degree to which reviewers unilaterally approve or decline new offerings. In particular, we would probably be most interested in whether there are appeal mechanisms or internal approvals necessary for a new offering to go forward. Since there is little evidence offered by Cooter, our analysis would be illustrative even if it were a small-n case study. In fact, a small-n case study may offer clues about the processes by which policy matters are delegated to low level technocrats, and thus provide openings for future research. Particularly in light of the problems making comparisons across political systems, we would probably want to develop our concepts before our empirics, à la Kurtz and Schrank.
Kurtz and Schrank conduct a lively critique of the measures developed by Kaufmann, Kraay and Mastruzzi (KKM)--their criticism of perceptual measurements of government effectiveness are persuasive; KKM's reply that they survey across many spreads of firms, IGOs, NGOs and individuals is a good defense, but ultimately one is left wondering if these different organizations' perceptions of different states are cross-fertilizing. The more substantive critique, however, are Kurtz and Schrank's questions about theoretical causality between governance and growth.
I wonder if the entire debate is being conducted too broadly: just as 'effective government' is a complex, wide concept, growth should be disaggregated as well. Would it not be more fruitful, in terms of research on how processes feed into one another, and for policy prescription, to posit hypotheses that specific elements of governance (e.g., educational investment) are linked to specific types of growth (e.g., increases in service industry income)? These more precise relationships--which relate the measurements more closely and intuitively to specific studies that have been done on those policies types--could then be aggregated, but we would have much more information than the broad brushstrokes being discussed here, where dependent variables seem to be limited to either short- or long-term growth of country GDP.
Shleifer and Vishny argue that corruption can take place under three environments: joint or monopolistic environment, a multiple independent monopolistic environment, and a competitive monopolistic environment. They go on to predict that the level of revenue collected in each (and the attendant distortion caused by the corruption) in respective descending order of magnitude is: the second case, the first case, and the last case.
ReplyDeleteOne way to test the theory is to leverage differences in the structure of regulatory responsibility across related, but distinct, regulatory areas and explore how corruption influences the ability of business people to operate under each. An excellent example can be found in a related work by Martin Dimitrov on intellectual property (IP) rights in China. Dimitrov classifies the regulatory scheme for patent, trademark, and copy right law as well as the ministries involved in enforcing each in China. Although he is interested in rule of law, he shows that each of these branches of IP law have differing numbers of regulatory agencies in charge of them, complicating enforcement in the case of trademark and copyright law where multiple agencies have responsiblity but simplifying it for patent law, where only one agency is involved. Using this as a base, one could then survey businesspeople in the related area in order to measure the level of distortion caused by corruption for each type of IP law (or whatever regulatory area one choose). Alternatively one could leverage cross-regional variation in local regulatory models for a single issue, such as building or business permits. This has the advantage of posing more controls, although it might be harder to conduct because one would need to classify regulatory procedures in a large number of local settings.
In either case, the measure of corruption would be the number of different agencies or groups one needs to speak with in order to obtain the regulatory good sought. This could then be used as an independent variable, along with controls for level of development, partisanship (in countries where it applies), etc. in a regression with the measure of corruption obtained from looking at the cross-regulatory (or cross regional) variation in the number of actors involved in obtaining the regulatory good.
This comment has been removed by the author.
ReplyDeleteMy apologies, I accidentally posted before completing the last paragraph. It should read:
ReplyDeleteIn either case, the measure of corruption would be the number of different agencies or groups one needs to speak with in order to obtain the regulatory good sought. This could then be used as an independent variable, along with controls for level of development, partisanship (in countries where it applies), etc. in a regression with the measure of distortion obtained through surveys of business people who are involved in that particular regulatory area. Possible instruments might include asking how much respondents believe corruption hampers their business, asking them to estimate the costs of such corruption, etc. (questions similar to this have been asked in the European Bank on Reconstruction and Development and World Bank's Business Environment and Enterprise Performance Survey in Russia). If this is the dependent variable, then one would also want to control for enterprise level features such as firm size, etc.
References (forgot to post them)
Dmitrov, Martin (2008). Piracy and the State: The Politics of Intellectual Property in China. Cambridge: Cambridge University Press.
Dmitrov, Martin (2004). Ph. D. Dissertation: Administrative Decentralization, Legal Fragmentation, and the Rule of Law: The Enforcement of Intellectual Property Rights Laws in China, Russia, Taiwan, and the Czech Republic. Stanford University.
Survival (duration) analysis would be one way to empirically test Geddes’ hypotheses on the effects of political competition on civil service reform. In this analysis, the dependent variable would be the timing of policy adoption. The individuals in the sample are countries, which are observed when they enter the sample and up to their exit from the sample or failure (here, countries fail when they decide to adopt civil service reforms). Among the independent variables, one would include factors such as: a) % of seats heald by the main parties in the legislature, or better still, b) the difference in seats between the incumbent’s party and the largest opposition party in the legislature, as a proportion of the total number of seats (see Murillo and Gallardo AJPS 2007 for such an empirical strategy). These measures would capture the relative power of parties in the policymaking process which is the key mechanism affecting the incentives of politicians to reform in Geddes’s model. Finally, a full set of controls are needed: these are not only political in nature (such as the ideological distance between parties, or the effective number of parties) but also economic (whether countries face sever fiscal constraints, and the rate of economic growth). Ideally, the sample would include all Latin American countries (not just 5 cases) and the time period could we extended to include the most recent democratization wave (this would fulfill the Geddes condition of studying 15 or more years of democratic stability).
ReplyDeleteAs a complement it would help the model a lot to find systematic relationships between the independent variables discussed above and blunt patronage measures such as the share of public employment in the economy or the contribution of salaries in total public expenditures. We could imagine Graphs that show how changes in the difference in seats between the largest parties are correlated or not with changes in patronage outcomes.
Geddes observes that the access to patronage is closely related to electoral system (i.e. PR system with an open or closed list). Reforms are possible and sustainable only when the top (two) parties possess similar legislative clout (access to patronage) and high party discipline. She concludes that administrative reform is difficult in fragmented party system. Both the party system and party discipline are endogenous to electoral rules. A follow-up question will be: under what conditions does the above equilibrium happen? When do the politicians need to depend on their personal votes instead of party votes? One condition referred in the Geddes piece is the PR system with an open list. Another condition I can think of is the multi-member-district under majoritarian rule, which often leads to intra-party competition.
ReplyDeleteOne implication from Geddes’ paper is that in a bi-party system with strong party discipline, administrative reform is most likely to be successful. Can anyone think of a counter-example?
I have posted a paper on courseworks.
ReplyDeleteI have posted my 2nd response paper on Courseworks.
ReplyDeleteComment Week 11 - Julie Touber
ReplyDeleteMarcus Kurtz and Andrew Schrank (KS) revisit the causality between institutions and economic growth. They see economic growth as the cause of good governance and not the opposite. In a response to KS, Daniel Kaufmann, Aart Kraay and Massimo Mastruzzi (KKM) defend the opposite: good institutions are the cause of sustained economic growth.
Identifying the relevant variables to evaluate good governance are debatable, and KS and KKM also engage in a debate on the perception of economic development, that according to KS is too limited to the business community. Timeline is also a key variable that is debated between KS and KKM. KKM point out the importance of long-term economic growth and reproach KS for looking at a very short period of time. It seems to me that KKM arguments are more plausible than KS’s one for this very last reason. However, the weakness of KKM argument is that it does not provide for an explanation of how some countries came to have good governance and from a policy point of view is quite pessimistic as it implies that no evolution is plausible.
One way to contribute to this debate and test parts of these arguments could be to look at foreign aid as “economic growth” and see which countries can sustain this growth or not, and for what reasons. I feel that absorptive capacity could be considered as a proxy for government effectiveness.
Cooter (2000) develops a delegation game to determine how a rational principal makes the decision to either directly exercise power or delegate it to an agent. The principal’s decision hinges on the probability of a good state of nature. Good luck can disguise diversion, while bad luck reveals it. As the probability of a good state of nature increases, the principal will be less likely to delegate. We could test Cooter’s prediction by examining the random event, the state of nature.
ReplyDeleteIn the game, the state of nature cannot be controlled or predicted by either the principal or the agent: “To illustrate, most state administrators cannot predict or control elections or the stock market” (Cooter 2000, 81). We could examine events that, in retrospect, would have changed the probability of a good state of nature (crisis, election, etc.) to determine if the principals in fact did change their level of delegation in response. We would expect that a shift to a good state of nature would decrease the level of delegation, while a shift in the opposite direction would increase the level of delegation. Ideally, we could use one probability-changing event across bureaucracies to control for the possibility that the nature of the event itself affects the principals' behavior. We would also have to control for the type of bureaucracy (if some are more sensitive to changes than others).
In Deliberate Discretion, John Huber and Charles Shipan attempt to develop a theory of how laws shape bureaucratic autonomy. They use legislative word count as their dependent variable. For instance in their example concerning the Medicaid program, they use as the dependent variable the change in the number of words in the statute in 1995-1996. There are a number of problems with this measure. First, word count may not necessarily be a good proxy for discretion. Looking at a statute in isolation may not give all the details concerning how discretion is delegated. A statute may refer to or incorporate another statute or existing regulations. Statutes in different policy areas or in different languages may just simply use differing amount of words. Legislators may also have different proclivities toward the number of words they have in a specific legislative proposal. These problems are attenuated when one uses changes in word count as a proxy for discretion. Huber et al’s measures does not accurately account for whether changes in the statute are due to varying levels of discretion or if they are due to some other variable. Although Huber et al. recognize this problem and attempt to control for it somewhat, there are still remaining problems with this measure. For instance, if a statute changes by 1000 words how does one differentiate between situations where the 1000 words were added or if the original statute contained 2000 words and 1000 words were deleted? Huber et al’s measure also self-admittedly acknowledges that use of word count as a proxy is very circumstance-dependent.
ReplyDeleteHow then can we create a better measure of discretion in order to test how laws shape bureaucratic discretion? This is a difficult problem and the alternative measures proposed suffer below from many of the same infirmities as the Huber et al. measure. Nevertheless, these alternatives are at least necessary to consider. Instead of focusing on statutes, one could focus on the scope, number and content of regulations. This, of course, relies on the assumption that administrative agencies have a custom of adopting regulations, as they do here in the United States. If this assumption is not true, then, this model could perhaps only be useful to measures discretion in places like the U.S., where regulations are freely promulgated. One could assess whether a greater number of regulations means that the agency has more discretion to promulgate its own rules with respect to the issue at hand. Although this too is a very-issue specific measure, it could perhaps be a better proxy for discretion. OLS or even negative binomial regression could be used to estimate the results. One could also measure bureaucratic discretion by creating an index. For instance, one could individually measure whether the bureaucracy has the ability to do certain things, i.e., they have the authority to promulgate regulations and how many of these regulations have been overturned, whether they have the authority to create commissions, etc. This index could also incorporate restrictions on the agency, such as whether the agency is subject to judicial review, sunset provisions, legislative oversight, etc. The codings on the various issues could then be summed up and each agency could be given a “score” on how much discretion it is given the legislature, on a specific bill, or collectively on many bills. There are still other possibilities for how we could measure discretion. One could measure it by the number of employees as agency has, speculating the discretion increases with the number of employees. One could also measure discretion by the amount of the agency budget, with the assumption that the greater the budget the more discretion the agency has.
Another problem with Huber et al’s measure is that it places primary importance on the legislature as the source from which discretion for the agency emanates. However, many agencies owe their allegiance to the executive branch, and thus, focusing solely on legislative instruments such as legislative word counts would not fully account for how much discretion the agency has. Thus, moving away from a legislative measure to a measure that focuses more on the agency itself (such as one of the measures suggested above) may paint a more accurate portrait of discretion, at least for these types of executive agencies. Of course, measuring the effect of law on discretion in executive agencies would require the analyst to employ different independent variables than the ones Huber et al. imply (though some, such as divided/unified government. etc could still be employed)
The debates between Kurtz & Schrank (KS) and Kaufmann et al. (KKM) focus on how to effectively measure “good governance”. KS criticizes KKM’s popular measures of government effectiveness (GE) being problematic because 1) they conflate policy preferences and outcomes with political institutions and 2) relies almost entirely upon the perceptions of businesspeople whose views of governance might be different from other groups of people. I think that the flaws of KKM’s measures can mostly be attributed to its subjective nature. The weaknesses of the subjective GE measure would be hard to eliminate even thought KKM had plausibly taken a number of modifying measures such as weighting the data collected from different sources/groups of people and framing the wording of survey questions as to assess the likelihood of governments’ expropriation rather than the actual expropriation etc. The subjective assessment of governance which mainly constitutes KKM’s GE measure has to more or less base on respondents’ previous experience to deal with the government and their existing perception of governance before the surveys were conducted. It is to be noted that people’s experiences and perceptions of governance are inevitably influenced by the economic performance of the time. As a result, KKM’s measure of governance effectiveness relates to previous economic growth by construction while it has been demonstrated to be important for future economic growth by “a large and careful” economic literature. No wonder other scholars including KS question the causality order between the two variables: good governance and economic growth. Temporal order between two variables is one of basic evidences by which we ascertain causality order between them. Since the subjective nature of KKM’s measures obscure the temporary order between good governance and economic growth, they are not helpful for us to sort out the causality arrow between these variables. In this sense, it is tempting to use objective measures of governance. Then the question becomes: can we objectively measure governance? If yes, what would the criteria of good governance be?
ReplyDeleteTo my knowledge, there are some quantitative and objective indicators measuring state capacity in existing literature and they might lend insight into measuring governance. For example, some scholars measure state capacity by the ratio of budgetary revenue/expenditure shares between the central and local governments (Wang and Hu, 2001). Because of the ratio form, the potential impact of the temporal economic performance on government’s budgetary or tax revenues can be controlled for. Because almost every government in the modern world has government finance statistics computed by relatively unified standard, such objective and quantitative measures of governance might be desirable in large-N cross-national comparison. Of course, whether objective measures like this can do a better job in measuring “a more basic notion of governance going back to the seminal work of Douglas North: the norms of limited government that protect private property from predation by the state” (KKM, 2007: 553) than KMM’s subjective one requires further validation.
Reference
Kaufmann, Daniel; Kraay, Aart; Mastruzzi, Massimo. "Growth and Governance: A Reply." the Journal of Politics 69, no. 2 (2007): 555-62.
Wang, Shaoguang; Hu, Angang. The Chinese Economy in Crisis: State Capacity and Tax Reform: East Gate Book, 2001.
Kurtz and Schrank argue that the understanding of the relationship between 'good governance' and economic growth has been made difficult by a general acceptance of the causal arrow going from the former to latter without conclusive evidence. What evidence there is, they claim, is potentially biased in that 'good governance' is defined from the point of view of businesses (whose interests may not align with that of citizens in general or of the state). They go on to argue instead that evidence seems to point to the inverse causal relationship: growth leads to better governance. One way to test this argument may be to begin at a national and sub national level (as the authors point out, perceptions of governance quality may be influenced by cultural perspectives, using these units of analyses would help eliminate this distortion). While it is difficult to come up with a universal and testable definition of what constitutes good governance, finding certain policies that would fall under such a category (though not be its full universe) shouldn't be that hard and may prove easier to test. The test that I am thinking of would then be to find a policy that can be defined as constituting a good governance practice that is articulated at the national level (this makes more sense in a federal system), and then to see whether implementation at the local/state levels precedes or follows any changes in economic growth at the subnational level. For example, beginning in the 90's in Brazil(following a national level articulation of the need to better State-citizen relations), state governments began creating government bodies that would unite in one location (per municipality) citizen access to information and services relating to the government (driver's licenses, government ids, applications for unemployment, pensions, criminal records, consumer complaints, etc.). The timing and extent of implementation and expansion of these bodies varied from state to state, with some beginning in 1991 while others as late as 2003. If economic growth positively affects good governance practices, can (preceding) growth at the local/state level account for the timing of the creation of these organs?
ReplyDeleteShleifer and Vishny (1993) employ the idea of modelling corruption as the outcome of demand and supply forces within an unofficial market for government goods and/or services and apply insights from industrial organization. They proceed to argue that corruption can take place in three different kinds of markets: monopolistic, competitive, or one with independent monopolists. Their mechanism for determining the effects of corruption in these various contexts focuses on the degree of coordination, or lack thereof, between public officials, whom private firms need to bribe to obtain licenses (or permits etc.) that are complementary to each other. They distinguish between two types of corruption that these officials may engage in: corruption with theft and corruption without theft. (Though I will not go into it here, it might be an interesting way to extend Shleifer and Vishny’s theoretical insights by exploring how they hold up in situations where public officials deal with corruption opportunities that are essentially unrelated).
ReplyDeleteShleifer and Vishny do not lend much empirical support to their claims, although they refer briefly to various historical cases. One way of testing their argument would be to look at levels of investment in countries with high levels of corruption vis-à-vis those with low levels. Based on their argument, we would expect to find that levels of investment – whether by the state itself, FDI, or by private actors within the country – would be lower in countries with high corruption, and thereby also find less technological innovation and perhaps even entrepreneurship, than in those with low levels. They cite Mauro’s (1993) finding that “holding 1980 real GDP constant, countries with higher corruption have a lower ratio of both total and private investment to GDP.” (p. 611). Further, they mention the problem and lack of foreign investment in post-Communist Russia and relate it to their model of a market with independent monopolists (p. 615). This example suggests the plausibility of their model. An updated and more rigorous empirical analysis, across a wide sample of countries, including data from the last 15 years (to the extent that this is available) would be a way of testing the model.
On a separate note, it might also be interesting to test Shleifer and Vishny’s arguments by examining entry and exit of firms in given sectors – construction, say, where government contracts, permits or licenses will need to be obtained by firms in order to operate, and could be subject to bribery or other forms of corruption – in countries with medium to high levels of corruption throughout the world (e.g. Nigeria, China etc.). Shleifer and Vishny themselves remark how a possible outcome of corruption with theft might be that firms who pay bribes will survive in any given industry or market and those who do not will go out of business (due to levels of costs, access to licenses, materials etc. essential to do business and so forth). We would therefore expect to see that firms that pay bribes in this context would be more likely to survive than firms that do not. Constructing an appropriate measure for this that captures the effect of corruption on firms’ fortunes might be difficult and it would certainly be a difficult task to establish causality (i.e. might other effects have an impact on the fortunes of the firm and not simply the impact of corruption? If so, which?). Still, it seems that this would be one fruitful way of empirically testing the implications of their argument, if an appropriate strategy can be devised.
Shleifer and Vishny’s propositions about corruption sit well with Cooter’s and Huber’s general arguments that relate institutional capacity to agency discretion and in turn corruption, which Shleiger and Vishny as well as Geddes argue effect economic development and growth adversely. As far as public sector services constitute input factors for private production, however, more efficient supply of government services through bribes might increase private sector productivity. An interesting question arises as to the extent to which the private sector surplus that arises from more efficient provision of public sector services might not only positively affect economic growth but also tolerate higher government taxation to offset losses due to corruption.
ReplyDeleteIn addition, to what extent might corruption be able to circumvent excessive regulation that itself was set up to increase corrupt rent-seeking by bureaucrats? More detailed models that study the effect of corruption across different levels of the bureaucracy might be able to detect when corruption by one agent (or in fact by the principal) at a higher level in the bureaucracy generates incentives for the emergence of corruption by other agents on lower levels of the bureaucracy? These models could advance our understanding of the effects of “corruption to evade corruption” on private sector output in developing and transitioning economies that feature complex corruption structures across several levels of the administration.
Shleifer and Vishny (1993) evaluate how government structure impacts the prevalence and nature of corruption. The conclusion of their analysis is that corruption can be reduced by political and economic competition – they argue that corruption is minimized where “several complementary government goods can be supplied by at least two government agencies” (606).
ReplyDeleteThis argument could be tested by examining a specific sector such as the construction industry across selected subregions and/or countries. The construction sector is particularly suitable because the ‘lumpy’ nature of investments, as well as the many bureaucratic permitting process make it especially prone to corruption. It is also a sector where one could envisage ‘multiple windows’ for government services such as building licenses and permits. To test the theory, the independent variable of interest would be the ‘plurality of licensing/permit processes” – i.e. a measure to determine the ‘competitiveness’, or number of government agencies in providing building permit and licensing approvals. The dependent variable would then be a measure of sector-specific corruption. This could be constructed along similar lines to the World Bank Governance index – comprising a mix of indicators of the perceived and measured probity and efficacy of government agencies relevant to the construction sector (Kaufman, Kraay and Mastruzzi, 2007). It is not clear that data is readily available on these measures. It is likely that they would need to be collected through field visits and interviews. Once date is collected however a regression analysis could be conducted to determine the correlation between ‘plurality’ and levels of corruption.
Cooter’s argument about the “delegation chain” is interesting, yet unsubstantiated by the type of empirical evidence that could lend it validity. His claim is that, as you move up the chain of power in a state’s administration (from the low-level bureaucrat to the top-level elected official), discretion increases. According to his analysis, since top officials often face a volatile policy environment, the costs associated with formal constraints would be prohibitively high to ensure effective policymaking. Furthermore, the level of scrutiny of top-level officials, in addition to their desire for reelection, reduces the incentive to be corrupt. For low-level bureaucrats, inflexibility is more appropriate since they face relatively stable policy environments, and are unmonitored by and unaccountable to the general public.
ReplyDeleteIt would be interesting to test this hypothesis in light of what we know about the technical expertise of lower-level bureaucrats. For policy issues where implementation is highly contingent on this technical expertise, would we really expect large constraints on low-level experts? In fact, on some highly technical policy matters, it seems that the low-level bureaucrats might actually have a higher degree of knowledge than the Secretary, or Minister of the relevant Department or Ministry. This would imply discretion increases as you go down the chain.
One approach to test this hypothesis would be to utilize the statutory analysis framework designed by Huber and Shipan in Deliberate Discretion? Because we are interested in looking at whether there is a higher level of discretion given to top-level officials than lower-level officials, we could examine regulations issued by departments, rather than statutes passed by a legislature. One promising area might be securities regulation, since independent regulatory bodies often govern it and it involves a continual and complex flow of technical information. We might expect that low-level bureaucrats would have a large degree of latitude in determining, for example, whether to approve new securities registrations. We could design an analysis of regulatory procedure across countries for new offerings by coding regulations by the degree to which reviewers unilaterally approve or decline new offerings. In particular, we would probably be most interested in whether there are appeal mechanisms or internal approvals necessary for a new offering to go forward. Since there is little evidence offered by Cooter, our analysis would be illustrative even if it were a small-n case study. In fact, a small-n case study may offer clues about the processes by which policy matters are delegated to low level technocrats, and thus provide openings for future research. Particularly in light of the problems making comparisons across political systems, we would probably want to develop our concepts before our empirics, à la Kurtz and Schrank.
Kurtz and Schrank conduct a lively critique of the measures developed by Kaufmann, Kraay and Mastruzzi (KKM)--their criticism of perceptual measurements of government effectiveness are persuasive; KKM's reply that they survey across many spreads of firms, IGOs, NGOs and individuals is a good defense, but ultimately one is left wondering if these different organizations' perceptions of different states are cross-fertilizing. The more substantive critique, however, are Kurtz and Schrank's questions about theoretical causality between governance and growth.
ReplyDeleteI wonder if the entire debate is being conducted too broadly: just as 'effective government' is a complex, wide concept, growth should be disaggregated as well. Would it not be more fruitful, in terms of research on how processes feed into one another, and for policy prescription, to posit hypotheses that specific elements of governance (e.g., educational investment) are linked to specific types of growth (e.g., increases in service industry income)? These more precise relationships--which relate the measurements more closely and intuitively to specific studies that have been done on those policies types--could then be aggregated, but we would have much more information than the broad brushstrokes being discussed here, where dependent variables seem to be limited to either short- or long-term growth of country GDP.