telecommunication sector in twelve countries located in Europe, Latin ... Flinders (2009) in a review article distinguish three different traditions in the study of.
Explaining the delegation of the telecommunication sector of twelve countries: An explanation using classification trees. Camilo Ignacio González and Koen Verhoest Introduction Research in regulation of liberalized markets traditionally has been focused on studying the delegation of regulatory competences from political principals (cabinet, ministers) to an independent regulatory agency based on the argument that it is a way to assure governments´ credible commitment to liberalization (Horn 1995; Yesilkagit 2004; Spiller, 1993; Gilardi, 2002). Hence, the emphasis of this research tradition has been very much on organizational aspects, with special attention for the characteristics of these sectoral regulatory agencies (like independence and regulatory competences). However, this organizational focus has its limitations. Particularly, when analyzing delegation patterns, existing studies do not distinguish between the different individual regulatory issues1 which are dealt with in regulatory decisions. Rather they tend to focus on broad categories of delegated decisions (e.g. the extent to which economic regulation is fully delegated to an agency or not) (see for example Gilardi 2008). However, regulatory decisions range from economic to social and technical regulation, and entail decisions about many different regulatory instruments (like licensing and authorization). Even in countries with very powerful sectoral regulatory agencies, not all individual regulatory decisions concerning economic, social and technical regulation of a market are delegated to such sectoral regulatory agencies, but are allocated to other actors or kept with the minister or its ministry. In this sense the question of which individual regulatory decisions are delegated and which are kept at the level of national ministers and ministries becomes relevant, as well as the question why this is the case. The basic argument of our paper is that when allocating regulatory competences to specific actors, legislators do indeed differentiate between different regulatory decisions. Moreover, they choose different delegation patterns for different individual regulatory decisions, following a certain criteria. Additionally, we argue that besides the former decision making logics, some other institutional factors at the country level or at the market level affect whether regulatory decisions are delegated or not. Based on the former, this paper aims to study, compare and explain delegation patterns at the level of individual regulatory decisions in the telecommunication sector in different countries with different administrative traditions. Thus, this paper aims to answer two research questions: The first one has a descriptive aim: What types of individual regulatory decisions are more likely to be delegated (away from the minister or ministry)?
1
By regulatory issues we mean the topics that are addressed in the regulation. We distinguish between regulation that aims to regulate economic issues, technical issues and social issues.
In addition, the second one has an explanatory aim: Why are some regulatory decisions delegated and others not? The main reason to study delegation is that it is process where, scholars and politicians, have put many expectations. Overman (2015), performed a literature review of 250 peer review articles about delegation. They found that there are three groups of expectations of politicians and researchers for delegation. The first set of reasons are economic. These expectations have their roots in public choice theory (Niskanen, 1971). Some arguments that rise in the economic expectations are increasing efficiency through competition and economies of scale (Bel & Warner, 2008); and the gain in efficiency by “letting managers manage” (Moon & deLeon, 2001). The second set of expectations are political in kind, these ones are more related with the delegation through agencification. In this group of expectations, one can find arguments such as the need to separate tasks execution from politics in order to assure objectivity and impartiality in public service delivery (Knox & Carmichael, 2006; Van Thiel & Yesilkagit, 2011). This depolitization of service delibery is expected to increase politcal credibility and political trust (Egeberg & Trondal, 2004; Bertelli, 2006). Additionally, it is expected that delegation will eacrease satisfaction with government and a more resposible public service deleibery (Van Thiel , 2001; Overman, 2015). Finally, the last set of expectation for delegatios are organizational expectations. In this type of expectations the central idea is that delegation will generate gains to the internal organization of public bodies. Here some arguments that stand out are: employees in decentralized units have a larger room for innovation and to work closely with their costumers (Peckham, Exworthy, Powell, & Greener, 2008). Additionally, there are some expectaiton related with an increase of public service motivation which will be aligned with the origanization´s mission (Brown, Potoski, & Van Slyke, 2006). Finally, there is the idea that through delegation organizations are more able to act in a flexible way (Bartelli 2005) and to deal with complex problems (Breul, 2010). The reason to conduct this research in the telecommunications market is that this sector was one of the first sectors to be liberalized worldwide, which is also the case for the countries that are included in this research. Therefore, the telecommunication sector is now quite mature, compared to other liberalized public services (e.g. post, electricity and public transport) (Bognetti & Obermann, 2008; Bitrán & Serra, 1998), and thus more delegation of regulatory competences may have occurred. Additionally, the matureness of the sector implies that delegation patterns in each country have had enough time to settle and therefore it is possible to find explanations that follow similarities or differences across countries. To answer the research questions we analyze individual regulatory decisions in the telecommunication sector in twelve countries located in Europe, Latin America and South Asia. To conduct the data analysis we use a classification tree model, which is a non-parametric statistical technique, which predicts the occurrence of a categorical dependent variable (i.e. decision been delegated or not) in terms of the classification power of different independent variables.
In the remainder of this paper, we will first present a discussion of the delegation literature, and then present the theoretical model and hypotheses that will guide our analysis. Then we present the methodological section followed by the results and discussion section. 1) Delegation: brief explanation of the concept. The study of delegation is quite common in political science and in public administration. Delegation can be understood as the transfer (fully or partially) of decision making competences from a public body, that is controlled by a principal that has been elected democratically, to another public body (Overman, 2015; Thatcher & Sweet, 2002). Delegation can take many forms: decentralization (Wallis & Oates, 1988; Rondinelli, 1981); agencification (Wettenhall, 2005; Verhoest, Thiel, Bouckaert, & Lægreid, 2012); and contracting out (Domberger & Jensen, 1997). In this paper, we discuss delegation in the context of agencification, the creation of specialized disaggregated bodies that perform specific tasks, which are in this paper, regulatory tasks. Flinders (2009) in a review article distinguish three different traditions in the study of delegation, which are rooted in different “disciplinary linages”. The first tradition is based on rational choice and transactional cost theory and is draws heavily on quantitative methods and on positivistic assumptions (Huber & Shipan, 2000; Epstein & O'halloran, 1999). This approach deals with issues such as explaining the circumstances of delegation, the strategic decisions made by agents when delegation occurs and the. The second tradition in the study of delegation is linked with the explanation of delegation within the European architecture. It draws on the previous tradition but it includes a new set of dependent variables and theoretical explanations. Thus, in this tradition according to Flinders besides rational choice and principal agent models this tradition includes new institutionalism explanations (Van Thiel , 2004; Yesilkagit & Christensen, 2009). Additionlly, this traditions, although, still uses deductive quantitative studies allows also detailed case studies and interviews. The third and final tradition on the study of delegation is more related with political studies as opposed to political sciences. Has a more inductive and qualitative approach and has a normative aim. In this tradition, critical studies, which aim to describe the politics of delegation are encountered (Weir, 1994). If the present paper had to be located in one the mentioned traditions it will fir mostly in the second traditions, as it includes both rational choice explanations and institutional expiations. 2) Theoretical framework In this section we present the theoretical framework that will guide the data analysis in this paper. As mentioned in the introduction we expect to determine why certain regulatory decisions are delegated and some other are kept at the ministerial level. Furthermore, we attempt to explain, which decision characteristics and which institutional characteristics affect the delegation of regulatory decision.
For that we will test different theories that try to explain why delegation occurs or why central, government decides to keep some regulatory decision to themselves. These theories will be nuanced by taking into account the characteristics of the regulatory decisions involved. The theoretical framework distinguishes four levels of explanatory factors:
Theories related to the type of regulation (economic, social or technical regulation), Theories related with the type of regulatory decision (strategic or operational) Theories at the country level (the level of veto players or political constraints), Theories at the sectoral or market level (the maturity of the liberalized market and the share of the public incumbent),
We will now elaborate the theoretical expectations that are going to be tested. These theoretical perspectives are discussed per explanatory level indicated above. Theories related with the type of regulatory decision Bureau Shaping Bureau Shaping is a theory first developed by Dunleavy (1991), and it was a response to rational choice models, and particularly to the theory of budget maximization. It criticizes mainly the assumption that bureaucrats will always try to maximize their bureau budget and size (Dowding, 1995). The basic idea is of the Bureau-Shaping model is that what ultimately is pursued by bureaucrats is to increase their bureau utility, and that this not necessarily requires an increase of size and budget. According to Dunleavy, bureaucrats are interested in two main sources of bureau utility. These are higher income and better work conditions (Dowding, 1995). The first source is increasing income through personal promotion, general salary rises, and individual contracts. The second source has to do with the type of functions that are assigned to the bureau, the location of the bureau and the atmosphere. This theory suggests that bureaucrats will tend to prefer strategic management and policy formulation tasks than day-to-day administrative functions. Additionally, in an economic application of the bureau-shaping model in the case of next step agencies in the UK, James (1995) proves that senior bureaucrats will give up budget increase in exchange of policy tasks. In this model, James suggests that there is a constrain to the increase of budget as this increases the time that bureaucrats have to spend in managerial work, and therefore at a given point bureaucrats will choose not to further increase their budget. In the case of regulatory decisions, we distinguish between two types of decisions, being either strategic or operational. Strategic decisions are those from a more general nature such as establishing regulatory guidelines, which apply to all regulatees, and the definition of the regulatory and telecommunication policies. Operational decisions, on the contrary, have to do with individual cases and have an individual application. Between those two types of decisions, we argue that, bureaucrats will prefer strategic regulatory decisions since they have a more policy-oriented nature.
Based on the bureau-shaping theory we can formulate the following hypothesis applied to our case: H1: Ministries will be less likely to delegate strategic regulatory decisions than operational regulatory decisions. The bureau shaping theory can be interpreted in a different manner, as the people that works in the sector regulators also can and provably will desire that strategic regulatory decision are delegated to them. Based on the former a second hypothesis could be formulated: H2: Sector regulators can influence the delegation process to receive more strategic than operational regulatory decisions. Therefore, it could be expected to see more strategic regulatory decisions delegated in comparison with operational regulatory decisions. Theories related with the type of regulation Before addressing the theories that are related with the type of regulation is necessary to explain what we mean be type of regulation. We distinguish between three types of regulation: technical regulation, economic regulation and social regulation. Technical regulation includes all the regulation that has to do with the distribution of limited resources and standards. In the category of economic regulation, we aggregate regulation that is related with the relation between telecommunication providers. Finally, everything that in relation with the interactions between providers and users refer to social regulation Credible commitment As mentioned in the introduction, credible commitment has been one of the major arguments that is used by scholars to explain the creation of a regulatory agency in a liberalized market. Additionally, credible commitment has been one of the main explanations used by scholars to explain delegation of regulatory decisions to regulatory agencies (Yesilkagit & Christensen, 2009; Baldwin & Lodge, 2012). However, we argue that this explanation may not hold for all type of regulation. Regulatory frameworks are composed with several types of regulation and not all of them are critical for investors trust. In the case of telecommunication, for instance, regulatory decisions that have to do with administrating the telecommunication fund or homologating telecommunication equipment (i.e. technical regulation), may not be as important for providers than regulatory decisions that have to do with interconnection between operators, market remedies or license granting (i.e. economic regulation). Furthermore, regulatory research has highlighted the fact that economic regulation is needed to assure the functioning of a competitive market, which is especially important when there is still a public incumbent (Coen & Heritier, 2007). Based on the former we argue that the credible commitment argument relates with issues like assuring competitiveness, avoiding market entry barriers, market abuse, competitive prices etc.
This type of regulatory decisions can be located in the above-mentioned economic type of regulation. Therefore, the following hypothesis is formulated: H3: Ministries will be more likely to delegate economic type of regulatory decisions than decisions that are of another regulatory type. Agency problem The agency problem is derived from the principal agent theory (Pratt and Zeckhauser 1991) which has its roots in economic transaction cost theory (Williamson, 1981). This theory stresses the difficulties in transactions and contractual relations between a principal and an agent, to whom the principal has delegated some decisions. Mainly the principal-agent problem is related with goal incongruency between both, and the asymmetry of information of the latter over the former and the tendency of agents to use that information to pursue their own interest. This theory predicts how and when central government (principal) will delegate certain specific functions over others to desegregated bodies (agencies). In this body of research we have identified two factors that may influence the delegation of regulatory decisions: decision characteristics based on rational choice theories and the political salience of the regulatory decisions -
Rational choice characteristics
These decision characteristics are drawn on a number on theories, which are based on rational choice theory: agency theory, transaction cost, and public choice theory. These characteristics are summarized by Verhoest et al. (2010). In their argumentation they present the characteristics that facilitates that the effectively delegation of tasks. In the case of this paper, we apply the same logic for delegation of regulatory decisions. Following these insights, we can argue that regulatory decisions are more efficiently delegated under the following conditions: when the results are easy to measure or observable, when objectives are easy to define (Jensen and Meckling 1976; Wilson 1989), when the involved activities are homogenous (Wilson 1989; Ter Bogt 1998), and when there is low level of assetspecific investments to be made (Williamson 1985; Van Thiel 2001). In our perspective, this type of decisions fit mostly in the abovementioned technical regulation. This type of regulation deals meanly with the access of operators to the spectrum, quality standards, equipment’s homologation, numbering and so on. Each of this decisions is rather homogenous and standardizable (e.g. same quality standards and procedures for all providers), their outcome is easy to measure (e.g. the use of a particular frequency) and there are low assetspecific investments to be made, as complexity is rather limited. Based on this we can formulate the following hypothesis: H4: Ministries will be more likely to delegate technical type of regulatory decisions than decisions that are of other regulatory type.
Political Salience Political salience is another concept that comes from agency theory that can be also linked with the delegation of regulatory decisions (Horn 1995; Yesilkagit 2004; Verhoest et al.2010). Political salience is high when the regulatory decision has a large direct impact upon a large amount of citizens or organizations. These types of decisions are likely to be more intensively controlled by the principals and, then, less likely to be delegated. In the field of telecommunications, these types of regulatory decisions, are part of the social type of regulation, and have to do with: user rights, universal service obligations, telecommunication investment projects and users complains. These type of regulatory decisions have the potentiality of affecting citizen’s support and voting behavior and hence involve more high risks for politicians in terms of re-election. Thus, politicians prefer to keep close control to them. In our case, highly political salient regulatory decisions are more likely to be found in the abovementioned social type of regulation. The reason for this is that social regulation has a more direct impact on how people perceive the service. Although, economic and technical regulation can and do affect the service people receive, social regulation deals with sensitive elements like billing, tariffs, and more important the right of each citizen to have access to telecommunication services. Therefore, the following hypothesis can be formulated: H5: Ministries will be less likely to delegate social type of regulatory decisions than decisions that are of other regulatory type Theories at the country level Political constraints (veto players) Currently there is a debate among scholars regarding the effect of political constraints on delegation. Two main accounts can be extracted form literature. The first account is based on delegation theory. Yesilkagit and Christensen (2009) argue that the creation of agencies and the particular characteristics of these agencies depend on policy conflict in a given country. This concept highlights the idea that the structure of the agencies reflects the particular political struggles and compromises that were on place at the moment of the agency creation. The argument suggests that in democracies it is not likely that the governing party can take all decisions on its own. Hence, opposition parties will try to influence agencies design and the regulatory decision that are delegated to these agencies (Riker, 1986). Even more, opposition parties will try to keep some influence over regulatory agencies after there are created. Therefore, we can expect that in moments of high political constraints founding legislation that creates agencies will create regulatory agencies to which rather limited amount of regulatory decisions are delegated. Moreover, in countries that experience high levels of political constraints we expect to find regulatory agencies with less regulatory decisions assigned.
Researchers have related policy conflict with the amount of independent veto player in a political system and with their preference (Yesilkagit & Christensen, 2009; Jordana & LeviFaur, 2005; Gilardi, 2002; Van Thiel, 2004). Veto players are understood as those actors that have the capacity to block policy change and can be located in the legislative branch (opposition parties), in the judiciary branch (judges who revise government decision) and on the executive branch (prime ministry, ministries, or ombudsman, or subnational levels of government). In the second account, some research (Keefer and Stasavage 2002, 2003; Moser 1999) have related political constraints and particularly veto players with lower levels of formal autonomy of regulatory agencies, particularly of central banks. In this account, veto players act as a functional equivalent for delegation. The idea is that if there are many veto players policies towards liberalization are not easy to change, thus, the need for credible commitment is lower, and there is no need to delegate regulatory decisions to an independent body. Based on this argumentation we argue that the level of political constraints has an effect on the delegation of regulatory decisions. Despite the fact that the two mentioned accounts differ in their explanation they both suggest a negative relation between policy conflict and delegation Thus, it is possible to specify the following hypothesis: H6: In countries with higher levels of political constraints it can be expected to observe lesser delegation of regulatory decisions to regulatory agencies, in comparison with countries with lower levels of political constraints. Theories at the sectoral or market level Market matureness The level of market matureness refers to a combination of two indicators, being the level of competitiveness of a market and the extent to which the incumbent is publicly or privately owned. This is based on the idea than after liberalization, a market is supposed to become more competitive (liberalization) on the one hand and to exhibit less public ownership regarding the main operators (privatization), on the other. The matureness of the market has a direct effect on the need for regulation. Particularly, regulation often occur as a consequence of market failures (Baldwin, Cave, & Lodge, 2012 Chapter 1). Theory of regulation, has suggested that the main reason to regulate in the need to correct market failures, such as negative externalities, incomplete information, entry barriers, uneven risk delivery, and monopoly abuse (Stiglitz, 2008). Additionally, scholars have also highlight the relation that exists between market competition and regulation. In that regard Jordana and Levi-faur (2004 pp.6) distinguish between regulation of competition and regulation for competition to describe the positive relation between competition and regulation. Both concepts assume a positive intervention of the state in economy but the latter implies larger capacities to do so than the former. The regulation for competition is normally done in sector specific bases and by a sector regulator (IRA) which acts proactively in a rather ex ante manner. Regulation of competition is done in a broader
manner (whole economic) and with less intrusive capacities in an ex-post manner. The latter is normally done by a general competition authorities. Coen and Heritier (2007) developed a conceptual model that tries to explain how regulation occurs in relation with the evolution of a given market. The basic argument is that the amount and intensity of the regulation needed has a direct relation with the moment in which a market is. When markets are less mature, the production of regulation is extensive and intense in order to stablish a market where there was not one before, and in other to prevent monopoly abuse (Bauer, 2005; Edwards and Waverman 2005). In this moment, thus, more regulatory decisions need to be allocated in the sector regulator. When markets start to become more mature and the incumbent has been fully privatized the chances of market failure are less and the need for extensive ex ante regulation is lower since with a mature market the need to create and ensure competition ex ante is reduced, thus less regulatory decisions delegated. Based on the former it is possible to generate the following hypothesis: H6: In markets that are less mature it can be expected to find more delegation of regulatory decision than in markets that are more mature. 3) Methodology In this section we present the main components of the methodology section, particularly we specify the country selection, how the data was collected, analyzed and how the cases were selected. Country selection For this paper, we have included twelve countries which are part of three different geographical regions and which have distinct administrative cultures and traditions (Painter and Peters 2010) namely (a) Europe: Belgium, Ireland, the Netherlands and Switzerland; (b) South Asia: Bangladesh, Nepal, India and Sri Lanka; and (c) Latin America: Colombia, Peru, Venezuela and Ecuador. The reason to select this country sample was to on the one hand to be able to see what general patterns could be found despite the heterogeneity of the cases and to see whether country-level variables like political constraints (which may be expected to be different across these countries) have any explanatory value. Data Collection The data collection for this paper was done in 2014. To conduct this research we used as main source for data collection the telecommunication laws of the twelve studied countries. When the primary telecommunication laws referred to secondary legislation or supranational directives, we also used this additional sources. From the legal instruments of each country, we extracted the regulatory decisions that are part of the regulatory framework. In order to limit the amount of decisions that were included and to make the results more easily comparable across countries we decided to allocate the decisions to a pre-defined set of issues. These issues,
involving each several regulatory decisions, were defined using a two-way strategy: first inductively after collecting data in twelve countries, we extracted some common issues that were included in the regulation of every country. Secondly, we compared and refined our list of issues with the issues highlighted in manuals for telecom regulation, published by the International telecommunication Union (ITU) (Blackman, 2011). This process resulted in a list of 14 issues. These issues however, were aggregated in three mentioned types of regulation, which match with the proposed hypotheses for the analysis. The issues and the aggregated categories are presented in table 1. Table 1: List of issues coded in the telecommunication legislations Issue Numbering Technical
Frequencies and spectrum Technical Standards Infrastructure building and share use of infrastructure Interconnection
Economic
License and general authorization Definition analysis and Market remedies (ex- ante remedies) Ex post regulation (based on general competition law) Universal access and social telecommunication policy (funds)
Social
Quality of service Users rights Tariffs and Billing
Additionally, we generate one extra variable on which we coded all regulatory decisions: the type of decision. This category captures if a decision is strategic or operational. This category is transversal to the different type of regulatory decisions listed in table 1, and were created for with the purpose of test the theoretical framework. For each decision, we mapped three elements: First, the number of decisions mentioned in the country legislation, the number of actors involved in each decision making process, and finally, the influence of each actor in each decision. The influence of each actor in the regulatory decision is based on set of influences scores presented in Table 2. We then transform the influence scores in order to code whether a decision was delegated or not. To do so we looked, for each country, at the influence scores of the Ministry or government for each individual decision and recoded in the following way: if the influences score was 0.6, 0.8 or 1, we coded the regulatory decision as being ‘not delegated’. This because the ministries or government remains with the capacity of defining the outcome of the regulatory decision, even if other actors are involved in the regulatory decision. Therefore, even when decision are shared with other actors, if ministers have veto power over the decisions we coded as a decision that is not delegated.
Consequently, we coded as a delegated decision those were the influence score of the ministry was 0.4, 0.2 or 0. In this case, the ministries or governments did not have the capacity of influencing the final regulatory decision outcome; even if they do provide some advice, this advice is not binding. Table 2 measurement of actors´ influence Weigh 0 0.2 0.4
Coding Not involved Informed Consulted
Description The actor is not involved in the decision The actor is informed about the decision The actor is consulted or gives a non-binding advice 0.6 Binding opinion The actor holds a binding position or can make proposals 0.8 Co-decision-maker The actor is a co-decision-maker 1 Main decision- The actor is the main decision-maker maker Source: Aubin & Verhoest (2014) 4.1.2) Data analysis The method of analysis used for this paper is a CART model (classification and regression tree). We use the classification tree model as opposed to the regression tree model, since it is the most appropriate in case one is working with categorical dependent variables. A classification tree is a method, which aims to predict an outcome based a series of binary splits of subsets of the data using for that the independent variables fitted in the model (Kurt, Ture, & Kurum, 2008). Thus, each independent variable will be divided in two, according to the membership or not of cases in a given value of the independent variable, with the aim of classifying the data in a way that only cases of the outcome that belong to the same category are put together. Classification trees start by choosing the best predictor, the one that has the largest explanatory power. This is defined by looking which predictor (independent variable) when split in two, causes the least misclassifications of the outcome (Rokach & Maimon, 2014). This main predictor is the starting point of the tree and it is call the root or the main node. After the root of the tree is defined, the model continues to split each independent variable according to their classification capacity. These variables will be labeled as intermediate nodes from which two branches develop. This process continues until all cases classified belong to the same category of an outcome, which is called a leaf, or until no further splits are possible (Rokach & Maimon, 2014). In the case of the classification tree presented in figure 1, each node represents an independent variable, each line represents a splitting criteria and the leafs represents a final classification, a situation in which all cases share the same category of the outcome. In the example presented in figure 1 the rood node is split in two, according to a given value if it is a continues variable or a given category if it an ordinal or nominal variable. Thus, all the cases which are covered by this value or which fall in this category will be classified to the left and those that do not will
be classified to the right. That process is further done with all the other variables that become secondary nodes, until a leaf is generated. As can be seen in figure 1, classification trees have a hierarchical nature. This means that the explanatory power of each independent variable is not considered independently as in more classic statistical techniques (linear, logit, or probit models), but in relation with the other predictors. For instance, in figure 1 the relevance to classify an outcome of a secondary node depends on how the data is classified in the root node. Figure 1 shows the basic structure of a classification three Root Node
secondary Node
Leaf
Secondary Node
Leaf
Leaf
Even though, it is not so commonly used in the social sciences in general or in public administration in particular, we believe that classification trees provide some very interesting analytic possibilities. There are some examples in the social sciences that use the classification tree model, in political sciences and law some examples are Ruger, Kim, Martin, & Quinn (2004) and Kastellec (2010). The use the classification tree model to predict decisions of judiciary bodies. In psicology this method has been used by Steadman, et al. (2000) and Monahan, et al. (2000) to preict violence risk in patince. This thecnique has been also used substantially in financial and economic research (Frydman, Altman, & Kao, 1985; Ngai, Hu, Wong, Chen, & Sun, 2011; Thomas, 2000) and in medicine (Arend, et al., 1990; (Lemon, Roy, Clark, Friedmann, & Rakowski, 2003). In the mentioned studies some common advantages that are mentioned regarding the classification tree approach are that it allows to combine the explanatory power of the different independent variables. This is particularly important since social phenomenon are seldom explained by independent effects but rather by a combination of factors. Hence, this type of analysis is more appropriate for the kind of analyses we would like to perform in this paper. Second, since classification trees are a non-parametric technique, there are no assumptions regarding distributions or variance. This makes a more flexible technique, which in turn is more suitable for working with nominal or ordinal variables, which is very common in the social sciences. Finally, classification trees are considered a rather transparent and easy to interpret technique, which is very common in data mining (Rokach & Maimon, 2014).
For the analysis in this paper, the dependent variable is a nominal variable with two values: no delegation coded with 0 and full delegation, coded with 1. This was done based on the recodification of the influence scores explained above. The independent variables for the analysis were coded as follows: concerning to operationalize the different types of regulation we created one categorical variable based on table 1. This variable has three categories: decisions related to technical regulation, decisions related to economic regulations, and decisions related to social regulation. We also created a dichotomous nominal variable, which distinguishes whether a decision is strategic or operational. To operationalize the independent variable at the country level, being the political constraints factor we use the political constraints index (POLCON V) (Henisz, 2000) which gives a measure of how easy a policy could be changed in one country. It measures the extent to which a change in preference of one actor may cause a policy change (Henisz, 2000). This index looks at the number of veto players in a particular country and the preferences of those veto players in a particular moment of time. The POLCON index was then transformed in a categorical variable with three categories: high political constraints, medium political constraints and low political constraints. The reason to do this was to generate more aggregated categories and to keep all the variables in the model either nominal or categorical. The independent variable that work at the market or sectoral level, being the market matureness, was operationalized by the creation of an index that measures its two dimensions: the level of competitiveness of the market and the extent to which there the incumbent is publicly owned. To do so, we used the data from the Little Data Book on Information and Communication Technology, a publication of the World Bank. This source provides information about the main fixed Line operators and the level of competitiveness of the main submarkets of the telecommunication sector (long distance, mobil, internet and international gateway). The Little Data Book provides annual information; hence, what we did was to calculate the average of the index for the years 2005, 2006 and 2007, so five years or more before the measurement of delegation of the regulatory decision. This was done in this way because we expect that governments need time to adjust the delegation of a regulatory decision in reaction to the observed level of market matureness. As changing the delegation of regulatory decisions needs a change in the legal framework and hence an intervention of parliament and legislative actors, we assume a time lag of several years. With the market matureness index, we did the same transformation than with the POLCON index. Thus, we ended with a three categories categorical variable. The overview of the variables included in the model and its operationalization are shown in table 2
Table 3 presents the basic theoretical model its operationalization and coding. Theoretical factor
Variable
Bureau shaping
Strategic vs Variable name: type of decision operational regulatory decision a nominal variable with two categories: strategic decision and operational decisions.
Credible commitment
Economic regulatory decision
Agency problem
Technical regulatory decision
Political Salience
Social regulatory decision
Political constrain
POLCON index
Variable operationalization
Variable name: type of reregulation nominal variable with three categories: economic, technical, and social regulatory decision.
Variable name: Polcon Ordinal variable with three categories: high, mid, and low POLCON scores
Market matureness
Market matureness Variable name: Market matureness index ordinal variable with three categories: high, medium, and low matureness scores
4) Results
In this section, we present the results of the analysis. We first present some descriptive results, and then we present the results of the classification tree model. Descriptive results In table 4, we preset the frequencies of each variable and how decision are spread across country . Regarding the latter it must be said that the data set includes 505 regulatory decisions. From that total 39 regulatory decisions are from Bangladesh, 46 from Belgium, 49 from Colombia, 59 from Ecuador, 37 come from India, 32 belong to Ireland, 40 to Nepal, 66 are from Peru, 31 from Sri Lanka, 15 from Switzerland, 31 from the Netherlands, and finally 60 come out Venezuela. If we look to the percentage that each country represents we can see that the each region of the world is rather balanced. The European countries tend to represent fewer regulatory decisions (except for Belgium) the south Asian countries are the middle and the South American countries are the ones with the largest amount of regulatory decisions. Regarding the independent variables, it is possible to see that in the type of regulation the largest category is referring to economic regulatory decisions (43%) followed by technical regulatory
decisions (35.2%) and social regulatory decisions (21.8%). The type of decisions is rather balanced: 52.9% are strategic decisions and 47.1% are operational decisions. When it comes to the POLCON index the largest category is countries with high POLCON values (39%), very closely followed by countries with low POLCON values (34.4%) and the smallest category being countries with medium POLCON values (24.8%). As to the variable market matureness, the largest category is mid-market matureness with 58.8%, which is significantly larger than the categories high market matureness and low market matureness that only have 25.5% and 15.6% of the regulatory decisions respectively. It must be mentioned, that the frequencies of the dependent variables are shown across countries, hence each frequency represents the amount of each category of each independent variable in all countries combined. Table 4 frequency of explanatory variables Variable
Variable operationalization
Frequency
Percentage
Type of regulation
Strategic decisions
267
52.9
Operational decisions
238
47.1
Economic regulatory decisions
217
43.0
Technical regulatory decisions
178
35.2
Social regulatory decisions
110
21.8
High POLCON
201
39.8
Mid POLCON
179
24.8
Low POLCON
125
34.4
High market matureness
129
25.5
Mid-market matureness
297
58.8
Low market matureness
79
15.6
Bangladesh
39
7.7
Belgium
46
9.1
Colombia
49
9.7
Ecuador
59
11.7
India
37
7.3
Ireland
32
6.3
Nepal
40
7.9
Type of regulation
POLCON index
Market Matureness index
Country
Peru
66
13.1
Sri Lanka
31
6.1
Switzerland
15
3
The Netherlands
31
6.9
Venezuela
60
11.9
Results of the tree model In this section, we present the results of the tree models. Figure 2 presents the results of the model that classifies wheatear or not a decision will be delegated or not. The bold labels in figure 2 represent variables of the model, which are located according to their classification power being either in the main or root node, or either in an intermediate note. The boxes in the bottom part of the figure are the leaves of the tree, where no further splits can be done and the final classification is presented. The information on the leaves represent the number of decisions that were correctly classified as well as the ones that were misclassified and the percentage of decisions that represents each leaf. Figure 2. Results of the classification tree model for the variable delegation.
For instance, the leaf of the far left was classified as decisions that are fully delegated. This leaf, represents 47% of the decisions from which 177 were correctly classified (decisions that were actually delegated) and 35 were incorrectly classified (decisions that the model classified as fully delegated, but in fact, they were not delegated). On the opposite side of the tree, the far right leaf was classified as decisions being not delegated. This leaf represents 13% of all the regulatory decisions from which 44 regulatory decisions were correctly classified (decisions that in fact were not delegated) and 15 were incorrectly classified (decisions that were actually delegated). The variable that has the largest classification power is market matureness, which is in the main or root node refers to a low or medium level of market matureness. From that point onwards any branch that goes to the left means that the decisions fulfill the condition that is expressed in the main node. On the contrary, any branch that goes to the right means that the decisions that are being classified do not fulfill the condition of this main node. The former means that, for instance, all the decisions that are classified to the left of the main node satisfy the condition of belonging to a telecommunications market that has low or medium levels of matureness and all that are classified to the right meet the opposite condition, in this case being in a telecommunications market that has highly mature. Table 3 presents the summary of the classification paths shown in figure 2. The first element important to mention about table 3, is that in all the classification paths, the number of decisions that were correctly classified is larger than the number of misclassifications. Additionally, more decision paths lead to delegation in comparisons with paths that lead to no delegation.
Table 3. Summary of classification paths. No
Path
Percentage Classification
Correctly Incorrectly classified classified
1
Medium or low market matureness – 47% Low or medium political constraints
Delegation
177
35
2
Medium or low market matureness - 12% high political constraints- Economic regulatory decisions
Delegation
45
8
3
Medium or low market matureness - 6% high political constraints – technical or social regulation - low market matureness
Delegation
27
7
4
Medium or low market matureness - 10% high political constraints –medium market matureness
No delegation 26
18-+
5
High market matureness - medium or 13% low political constraints
No delegation 44
15
6
High market matureness - high 11% political constraints - technical or economic type of regulation
Delegation
2
7
High market matureness - high 1% political constraints - social type of regulation
No delegation 3
48
1
From all the paths presented in table 3, path 1 is the one that explain the largest share of regulatory decisions (47%). Paths 2, 4,5 and 6 have a similar explanatory power (around 10%) and paths 3 and 7 have the smallest explanatory power. This allows to make assess the empirical relevance of the different delegation paths. The first path is a combination of the two variables with the largest classification power: Market matureness, and political constraints. This path provides evidence that suggests that a less mature state of the telecommunication market, in a country with a low level of political constraints (low number of veto players) explain a great deal of why decisions are delegated, away from the minister or ministry. The second path shows that medium to low market matureness in combination with regulatory decisions that are of an economic type, also lead to delegation. This suggest that when markets are less mature, it is more important and necessary for government to signal credible commitment to investors, leading to the delegation of regulatory decisions. The third and fourth path shows that low market matureness is an especially relevant factor. These paths suggested that when a decision is not related to economic regulation (thus, being either social or technical) in case the market has a low matureness this will lead to delegation. Moreover in case that the market does not have a low level of market matureness, this would lead to no delegation. The fifth to seventh path presents the situation when the market has a high level of matureness (variable with the highest explanatory power). The fifth path suggests that high market matureness and high political constraints (high policy conflict, many veto players) will lead to no delegation. The sixth delegation path, suggest that in case market matureness is high and political constraints are high, technical or economic regulatory decisions will be delegated. The latter gives some support for the idea of the agency problem logic and the credible commitment
explanatory factor. On the contrary the model suggested in the seventh path states that in case the decision is not of a technical or economic nature (thus being social) there will be no delegation of such decisions. This path gives some credibility to the political saliency explanatory factor. However, this last path only covers 1% of the decisions. Discussion In table 4, we bring back the proposed hypotheses and discussed in light with the findings of this paper. The results of this paper, summarized in table 4 suggest that the level of the market is indeed the factors with the highest classification power. In forth paths low or mid-market matureness is associated with delegation, and in two cases, high market matureness is related with no delegation. Of course, this effect is nuanced by the other explanatory factors. The findings in this article about political constrain present an interesting insight regarding the role about veto players. As mentioned in the theoretical framework some scholars (Yesilkagit and Christensen, 2009; Shepsle, 1992; Moe, 1995; Jordana & Levi-Faur, 2005; Gilardi, 2002; Van Thiel, 2004) have related veto players with high policy conflict, and this is related, in turn, with less delegation. Some other scholars (Keefer and Stasavage 2002, 2003; Moser 1999) have suggested that veto players can act as a functional equivalent for delegation and thus many veto players are related with less delegation. Our findings suggest that political constrain, represented in the amount of veto players, can or cannot have a negative relation with delegation, as suggested by the literature, depending on the combination of the amount of veto players with other factors. Particularly the effect of political constraints seems to depend especially on the level of market matureness in each specific telecommunication sector of each country. When lower levels of political constrain are present in a country that has a telecommunication market that is not very mature, delegation of regulatory decisions is seen (path 1). This suggest that when markets are not much mature, the lack of veto players generates delegation. In this case, the finding go according to the theoretical expectation. Nevertheless, when low levels of political constraints happen in a country that has a more evolved market, it can be seen rather no delegation (path 5). In this case, few veto players so low policy conflict is related with less delegation. This goes against what literature suggest, and can suggest an interesting line for future research. The relation between the level of market matureness and the political constrain is nuanced by the type of regulation the decision and particularly by the credible commitment idea, which is related with economic type of regulation. Path 2 and 6 shows that even when the political constraints are high if the decision is economic that will lead to delegation. This suggest that the credible commitment need of governments is important even if political constrain is high (many veto players). Hence, when the regulatory decision has is from an economic type, high political constrain do not generate less delegation.
Table 4 summary of hypotheses and findings Hypothesis
Finding.
H1:Ministries will be less likely to delegate strategic regulatory decisions than operational regulatory decisions
The classification tree model did not show any support for this hypothesis. Furthermore, the variable “type of decision” was not included by the model to classify the regulatory decisions.
H2: it could be expected to see more strategic regulatory decisions delegated in comparison with operational regulatory decisions
Similarly than hypothesis 1, in The case of this hypothesis there are no evidence that gives any support.
H2:Ministries will be more likely to delegate economic type of regulatory decisions than other type of regulatory decisions
Economic regulatory decisions appear in two paths that lead to delegation (paths 2 and 6), this covered 33% of the decision. This type of decisions appear however, in combination either with low or mid-market matureness or with high political constrain.
H3: Ministries will be more likely to delegate technical regulatory decisions than other type of decisions.
This type of regulation only appeared in one path that leads to delegation (path 6). It coverers 11%. However, in that path technical type of regulatory decisions go along with economic type of regulatory decisions.
H4: Ministries will be less likely to delegate social type of regulatory decisions than other type of regulatory decision
There is little support for this hypothesis. Social type of regulatory decision only appear in path 7, which covers 1% of the regulatory decisions. However this path suggest high market matureness, high political constrain will lead to no delegation when the decision is of social type.
H5: In countries with higher levels of political constrain it can be expected to see lesser delegation of regulatory decision to regulatory agencies, in comparison with countries with lower levels of political constrain.
This is one of the hypothesis with largest support, the first path, that covers 47% of the regulatory decisions show that mid to low market matureness followed with low or middle political constraints lead to delegation. Thus less political constrain more delegation. However, paths 5 and 6 suggest give contradictory findings for this hypothesis. Path 5 shows that when markets are highly evolved, low political constraints leads to no delegation. Thus, low political constrain less delegation. This covers 13% of the decisions. Path 6 suggest high levels of political constrain leads to delegation when market are more evolved and the type of regulation is economic or technical. This covers 11% the decisions.
H7: In markets that are less mature it can be expected to see more delegation of regulatory decision than in markets that are more mature.
This is the hypothesis with more support, as the market matureness proved to be the theoretical factors with the largest classification power. Path 1, 2, 3 and 4 shows that low and mid-market matureness, in combination with low political constrain and the different types of delegation also leads to delegation. Additionally, path 5 and 7 suggest that when the market is mature and goes along with low political constrain or social type of regulatory decisions this will lead to no delegation. Path 6 however, suggest that when high market matureness is mixed and low and mid political constrain will lead to delegation. So in this case high market matureness is related with delegation of regulatory decisions.
The relation between credible commitment (economic type of regulation) and delegation, and the lack of relation between political constraints and credible commitment, go in the same direction than the finding of Gilardi (2003). Gilardi found out in a study than delegation was more related with economic regulation than with other types of regulation and that veto players did not have a negative impact on the necessity of credible commitment. What was been shown in this paper has some limitations: the first one of course is that it only covers the telecommunications sector; this implies that the delegation dynamics may not be the same in other sectors. Therefore, comparative research that looks to different utility sectors is needed. Additionally, this paper has data of twelve countries. It may be necessary to expand the number of countries and cover other regions. Finally, we looked in this article to delegation in a formal manner; we do not say anything about how delegation occurs in more “de facto” bases. More research will be welcome in that regard too. Despite the former, what has been presented here has many strengths. The first one is that we present data of countries in different regions; this means that what we found has significance, across different regions of the world. Second, in this paper we look at the regulatory decision level. Normally, research on delegation is seen at the sectoral regulator level and its level of independence. Therefore, we believe than looking at the decision level can give a complementary a more detailed account of the delegation process. Third, by looking to the different types of regulation, and types of regulatory decisions, we are able also to approach to the theoretical arguments in a more refined manner. Finally, this paper has a major methodological strength. Due to the analysis technique that was used we were allowed to go beyond the assessment of individual effects and see how the different variables used in this research interact with each other generating different delegations paths. This is interesting, as seldom in the social sciences phenomena has a single independent cause, rather what is found is that phenomena is due to the combination of different explanatory factors. The classification tree model allows us to capture that peculiarity. References Arend, W. P., Michel, B. A., Bloch, D. A., Hunder, G. G., Calabrese, L. H., Edworthy, S. M., . . . Zvaifler, N. J. (1990). The American College of Rheumatology 1990 criteria for the classification of Takayasu arteritis. Arthritis & Rheumatism, 8, 1129-1134. Aubin, D., & Verhoest, K. (. (2014). Multi-Level Regulation in the Telecommunications Sector: Adaptive Regulatory Arrangements in Belgium, Ireland, The Netherlands and Switzerland. Hampshire: Palgrave Macmillan. Baldwin, R. C., & Lodge, M. (2012). Understanding regulation: theory, strategy, and practice. New York: Oxford University Press. Baldwin, R., Cave, M., & Lodge, M. (2012). Understanding regulation: theory, strategy, and practice. New York: Oxford University Press.
Bauner, M. W. (2005). Administrative cost of reforming utilities. En D. Coen, & A. Héritier, Refining Regulatory Regimens (págs. 53-88). Cheltenham: Edward Elger Publishing . Bel, G., & Warner, M. (2008). Does privatization of solid waste and water services reduce costs? A review of empirical studies. Resources. Conservation and Recycling, 52(12), 1337-1348. Bertelli, A. M. (2006). Delegating to the Quango: Ex ante and ex post ministerial constraints. Governance, 229-249. Bertelli, A. M. (2006). Governing the quango: An auditing and cheating model of quasi-governmental authoritie. Journal of Public Administration Research and Theory, 16(2), 239-261. Bitrán, E., & Serra, P. (1998). Regulation of privatized utilities: the Chilean experience. World Development, 26(6), 945-962. Bognetti, G., & Obermann, G. (2008). Liberalization and privatization of public utilities: origins of the debate, current issues and challenges for the future. Annals of Public and Cooperative Economics, 79(3), 461-485. Breul, J. D. (2010). Practitioner's perspective—Improving sourcing decisions. Public Administration Review, 70, 193-200. Brown, T. L., Potoski, M., & Van Slyke, D. M. (2006). Managing public service contracts: Aligning values, institutions, and markets. Public Administration Review, 66(3), 323-331. Coen, D., & Héritier, A. (2005). Refining Regulatory Regimen. Cheltenham : Edward Elgar Publishing. Colin Hay, & Wincott, D. (1998). Structure, Agency and Historical Institutionalism. Political Studies, 951-957. Domberger, S., & Jensen, P. (1997). Contracting out by the public sector: theory, evidence, prospects. Oxford review of economic policy,, 13(4 ), 67-78. Dowding, K. (1995). The Civil Service . New York : Routledge. Egeberg, M., & Trondal, J. (2004). Governance. Political leadership and bureaucratic autonomy: Effects of agencification, 4, 673-688. Epstein, D., & O'halloran, S. (1999). Delegating powers: A transaction cost politics approach to policy making under separate powers. Cambridge: Cambridge University Press. Flinders, M. (2009). Review article: Theory and method in the study of delegation: Three dominant traditions. Public Administration, 87(4), 955-971. Frydman, H., Altman, E. I., & Kao, D. L. (1985). Introducing recursive partitioning for financial classification: the case of financial distress. The Journal of Finance, 40(1), 269-291. Gilardi, F. (2002). Policy credibility and delegation to independent regulatory agencies: a comparative empirical analysis. Journal of European Public Policy,, 9(6), 873-893.
Gilardi, F. (2003). Delegation to independent regulatory agencies in Western Europe: a crosssectional comparison. In University of Lausanne, Paper prepared for the workshop Delegation in Contemporary Democracies ECPR Joint Sessions of Workshops, 29, pág. 38. Edinburgh. Henisz, W. J. (2000). The Institutional Environment for Economic Growth. Economics and Politics, 1, 1-31. Huber, J. D., & Shipan, C. R. (2000). The costs of control: Legislators, agencies, and transaction costS. Legislative Studies Quarterly, 25(1), 25-52. Kastellec, J. P. (2010). The statistical analysis of judicial decisions and legal rules with classification trees. Journal of Empirical Legal Studies, 2, 202-230. Knox, C., & Carmichael, P. (2006). Bureau shuffling? The review of public administration in Northern Ireland. Public administration, 84(4), 941-965. Kurt, I., Ture, M., & Kurum, A. T. (2008). Comparing performances of logistic regression, classification and regression tree, and neural networks for predicting coronary artery disease. Expert Systems with Applications. 34(1), 366-374. Lemon, S. C., Roy, J., Clark, M. A., Friedmann, P. D., & Rakowski, W. (2003). Classification and regression tree analysis in public health: methodological review and comparison with logistic regression. . Annals of behavioral medicine, 26(3), 172-181. Megginson, W. L., & Netter, J. M. (2001). From state to market: A survey of empirical studies on privatization. Journal of economic literature, 39(2), 321-389. Meyer, J. W., & Rowan, B. (1977). Institutionalized Organizations: Formal Structure as Myth and Ceremony. American Journal of Sociology, 340-363. Monahan, J., Steadman, H. J., Robbins, P. C., Silver, E., Appelbaum, P. S., Grisso, T., . . . Roth, L. H. (2000). eveloping a clinically useful actuarial tool for assessing violence risk. The British Journal of Psychiatry, 176(4), 312-319. Moon, M. J., & deLeon, P. (2001). Municipal reinvention: Managerial values and diffusion among municipalities. Journal of Public Administration Research and Theory, 11(3), 327-352. Ngai, E. W., Hu, Y., Wong, Y. H., Chen, Y., & Sun, X. (2011). The application of data mining techniques in financial fraud detection: A classification framework and an academic review of literature. . Decision Support Systems, 50(3), 559-569. Niskanen, W. A. (1971). Bureaucracy and representative government. Chicago : Aldine Atherton . Overman, S. (2015). Great Expectations of Public Service Delegation: A systematic review. Public Management Review, 1-25. Peckham, S., Exworthy, M., Powell, M., & Greener, I. (2008). Decentralizing health services in the UK: a new conceptual framework. Public Administration, 86(2), 559-580.
Pierson, P., & Skocpol , T. (2002). Historical Institutionalism in Contemporary Political Science. In I. Katznelson, & H. Milner, Political Science: The State of the Discipline (pp. 693-721). New York: W.W Norton. Rokach, L., & Maimon, O. (2014). Data mining with decision trees: theory and applications. Singapor: World scientific. Rondinelli, D. A. (1981). Government decentralization in comparative perspective theory and practice in developing countries. International review of administrative sciences, 47(2), 133-145. Ruger, T. W., Kim, P. T., Martin, A. D., & Quinn, K. M. (2004). The Supreme Court forecasting project: Legal and political science approaches to predicting Supreme Court decisionmaking. Columbia Law Review, 1150-1210. Steadman, H. J., Silver, E., Monahan, J., Appelbaum, P. S., Clark Robbins, P., Mulvey, E. P., . . . Banks, S. (2000). A classification tree approach to the development of actuarial violence risk assessment tools. Law and Human Behavior, 24(1), 83-100. Stiglitz, J. E. (2008). Government failure vs. market failure: Principles of regulation. Thatcher, M., & Sweet, A. S. (2002). Theory and practice of delegation to non-majoritarian institutions. West European Politics, 25(1), 1-22. Thomas, L. C. (2000). A survey of credit and behavioural scoring: forecasting financial risk of lending to consumers. International journal of forecasting, 16(2), 149-172. Van Thiel, S. (2001). Quangos: trends, causes and consequences. Aldershot: Ashgate. Van Thiel, S. (2004). Trends in the Public Sector Why Politicians Prefer Quasi-Autonomous Organizations. Journal of Theoretical Politics, 16(2), 175-201. Van Thiel, S., & Yesilkagit, K. (2011). Good neighbours or distant friends? Trust between Dutch ministries and their executive agencies. Public Management Review,, 13(6), 783-802. Verhoest, K., Roness, P. G., Verschuere, B., Rubecksen, K., & MacCathaigh, M. (2010). Autonomy and Control of State Agencies. New York: Palgrave. Verhoest, K., Thiel, S. V., Bouckaert, G., & Lægreid, P. (2012). Government Agencies Practices and Lessons from 30 Countries. Basingstoke: Palgrave Macmillan. Wallis, J. J., & Oates, W. E. (1988). Decentralization in the public sector: An empirical study of state and local government. En H. S. Rosen, Fiscal federalism: Quantitative studies (págs. 5-32). Chicago: University of Chicago Press. Weir, S. (1994). Ego trip Extra-governmental organisations in the United Kingdom and their accountability; democratic audit. London: Charter 88 Trust. Wettenhall, R. (2005). Agencies and non-departmental public bodies: The hard and soft lenses of agencification theory. Public Management Review, 7(4), 615-635.
Williamson, O. E. (1981). The Transaction Cost Approach. American Journal of Sociology, 87(3), 548577. Yesilkagit, K., & Christensen, J. (2009). Institutional Design and Formal Autonomy: Political versus Historical and Cultural Explanations. Journal of Public Administration Research and Theory, 53–74.
