Good Governance what we think it is and what we really measure.

Good Governance what we think it is and what we really measure. Björn Halleröd 1, Hans Ekbrand 1, and David Gordon 2 1 Department of Sociology and Work Science University of Gothenburg Sweden 2 School for Policy Studies University of Bristol UK Early draft presented at The Quality of Government and the Performance of Democracies seminar, Gothenburg, May 20-22, 2014 Abstract The so-called institutional revolution within social science is closely related to the use of a broad range of governance indicators provided by a variety of organizations. However, it is from a theoretical perspective often far from clear what these indicators are supposed to measure, which also makes it hard to draw concise policy conclusion. We use a structural equation (SEM) approach to analyze a broad range of governance indicators and we draw the following conclusions. Exiting indicators cannot in a straightforward application distinguish theoretically derived aspect of governance. They are mainly measuring one single latent factor, a factor we at this stage label as quality of government (QoG). However, QoG appears not to be measuring a unique feature but is more realistically to be interpreted as an approximation of GDP per capita. Because we use SEM we have been able to modeling the error terms and extracting a residual latent factor. The residual factor captures the degree of democratization and existence of liberal human rights net of QoG. The residual factor makes it possible to analyze if democratization and liberal rights impact on for example living conditions, access to education et cetera net of QoG (which is to be interpreted net of general economic development). Our preliminary analysis shows that democratization and liberal rights are unrelated to the existence of basic social rights. Hence, the analysis supports earlier findings that indicate that democratic values and institutions per se have little impact on policy outcomes and peoples living conditions. However, our results strongly advice against analyses that uses different indicators of governance in order to determine what aspects that are most important given a certain outcome. 1

Introduction During the past decades we have witnessed the institutional revolution and social scientists has at least since early 1990s and onwards stressed the importance of certain types of rules and regulations for the wellbeing of societies (North, 1990; North et al., 2009; Ostrom, 1990; Holmberg and Rothstein, 2012; Rothstein, 2011; 1998; Dellepiane-Avellaneda, 2010). At the same time, or as a consequence, a large majority of all developmental institutions are today promoting quality of government (QoG) as a central part of their agenda (Langbein and Knack, 2010; Gisselquist, 2012; Arndt and Oman, 2006). The development has been promoted and facilitated by a large number of measures of different aspects of QoG developed by researchers and organisations such as the World Bank, Transparency International, and Freedom House. In this article we will analyze if it is, given existing indicators, at all possible to empirically distinguish different theoretically derived aspects of QoG. The aim is to investigate if available and frequently used measures of QoG are measuring different theoretically meaningful and empirically valid aspects of QoG or if they in fact are measuring the same thing over and over again. If the former is true, countries can be meaningfully ranked on different QoG dimensions and we can judge each dimensions relative importance for outcomes such as economic growth, child poverty, or educational achievements and, not least important, we can let the results inform policy choices. If the latter is true we can still rank countries and relate the ranking to outcomes but we cannot not empirically distinguish between different aspects of QoG and, hence, not formulate distinct policy advices. Background Since the mid-1990s, evidence that corruption and other forms of poor quality of government (QoG) constitute a general social ill has been mounting and there is a broad consensus that good QoG is a pre-requisite for long term and sustainable increase in living standards (Dellepiane-Avellaneda, 2010). Building on seminal works of Douglas North and Mancur Olsen new institutional economists have underlined the importance of well functioning institutions for economic growth. Aspects such as rule of law, property rights, political stability, and lack of corruption has been pointed out as essential in order to promote economic growth (Dellepiane-Avellaneda, 2010; Voigt, 2012). Given the assumed impact 2

of QoG on economic growth it is easy to see the importance of QoG for increase in living standards. But, as it seems good QoG does not only promote economic prosperity. There is also broad support for the idea that QoG, net of economic growth, has strong direct implications for population health, people s access to basic services such as safe water, healthcare and education, economic equality, social trust, political legitimacy, intra-state as well as inter-state stability, and people s subjective wellbeing (Halleröd et al., 2013; Holmberg and Rothstein, 2012; Rothstein, 2011; Uslaner, 2008; Tavits, 2008; Svallfors, 2013). That is, QoG offers an explanation to why living conditions vary dramatically also between countries with approximately the same GDP per capita. In many case, mainly among economists, democratic institutions are seen as an aspect of QoG whereas others, predominantly political scientists, often wish to separate democratization from QoG. Analysis following the latter tradition often shows that QoG variables have substantially higher positive correlations with both economic growth and standard measures of human well-being than do measures of democracy (Dellepiane-Avellaneda, 2010; Rothstein, 2011; Sen, 2011; Przeworski et al., 2000). The conclusion to be drawn is that even though democracy can be seen as a good thing, it will not promote growth or eradicate poverty if the wider system of governance is incapable of, in a first step, formulating functioning polices and, in a second step, secure implementation in a trustworthy, predictable and impartial manner (Rothstein, 2011). What is often less clear in the literature is what kind of mechanisms link QoG to economic and social development, how to understand the direction of causality, and what aspects of QoG sound policies, efficient administration, control of corruption, etc. are of most importance. An even more fundamental question is if it is QoG or the actual policies that are most important and the causal relation between policies and QoG (Glaeser et al., 2004). QoG what is the problem? An example of broad theoretically based critique of how QoG has been conceptualized and measured is provided by Agnafors (2013). Criticizing existing definitions of QoG for being to narrow and atheoretical he argue for a definition that include moral content and encompasses a plurality of values and virtues at its core (Ibid p. 433). This type of argumentation certainly has its value clarifying the complexity of QoG and pointing at weaknesses in existing approaches but it also has, apart from the 3

sheer complexity and the fact that general acceptance for a settled moral content is far from given (Rothstein, 2014), fundamental shortcomings. First, the suggested definition does not distinguish between how institutions are organized (the rules of the game) and outcomes (how the game is played) making it more or less impossible to draw causal conclusions, which is something that has been discussed in length by among others Glaeser et al. (2004). Second, the inclusiveness of the suggested definition conceals the relative importance of the different components, which has detrimental effects on the usefulness of the definition (Gisselquist, 2012). The policy relevance of this type of encompassing definition is therefore highly questionable. However, even though we do not agree with Agnafors concerning the conclusions, we do argue that today s measures of QoG in fact include moral content and encompasses a plurality of values and virtues at its core. The difference is that we perceive this as a major problem, not a solution. As pointed out by Gisselquist (2012) almost all major development institutions underline the importance of QoG. However, as well illustrated good governance means different things to different organizations and different actors within these organizations (ibid: 0). Pointing the way forward Gisselquist, correctly on our view, argues for a strategy that is theoretically informed and draws on accumulated academic knowledge. However, taking the step from theoretical clarity to actual knowledge requires empirical indicators that in a reliable way can distinguish different components of QoG. The problem we want to point at is that we doubt that available indicators of QoG provide the empirical information necessary to do that. In a seminal article, Glaeser et al. (2004) discuss whether institutional quality, understood as constraints on executive power, causes economic growth or if it is the other way around, i.e., as countries got richer they tend to build better institutions. The overall conclusion drawn by Glaeser and his colleagues is that growth precedes institutional quality, including democratic development, and that growth mainly is driven by accumulation of human capital. But, how does human capital evolve? Today there are plenty of evidence of the detrimental effect of corruption and lack of administrative capacity on the educational system (Banerjee and Duflo, 2011). Hence, if good QoG is needed for the development of human capital via the educational system it seems odd to argue for a causal chain that sees human capital as a more or less exogenous factor that first causes growth and thereafter good 4

QoG. We have to ask ourselves why some countries manage to raise the level of human capital. If it is not good governance, what is it? The issue of causality is central, as we need to know what aspects of QoG cause what. Only then can research provide scientifically based advised about what needs to be done in order to improve peoples living conditions. Strong correlations are often an asset when the order of causality is given. We can be sure that smoking causes lung cancer because it is unlikely that lung cancer causes smoking. We can test if recurrent earthquakes prevent countries from building functional institutions because it is unlikely that bad QoG causes earthquakes (Daoud et al., 2014). However, even though correlations are strong the nature of the causal relationships between different aspects of QoG, economic growth, formation of human capital etc. is highly contested. In fact, the inconclusive results concerning causal relations might indicate that there are no unidirectional dominating causal relationships to be found. In the real world it might be that building of institutions, economic development, democratization, and formation of human capital are deeply interweaved and that differences between countries best are understood in terms of good and vicious circles and that the way events are ordered within these circles are more or less unique for each country. This reasoning point towards a methodological approach that is less focused on disentangling specific causalities and more focused on understanding what combination of circumstances are necessary for a country to enter a good circle. If we for example look at how countries scores on the World Bank Institute s (WBI) measure of governance efficiency we can see that all top scoring countries belongs to the OECD, i.e., they are at least in economically terms the most developed countries. However, when looking at how governance is organized within this countries we find a large degree of variation (Andrews, 2008). In fact, it is the causes and consequences of this variation among the top scoring nations that is the main object for the vast comparative welfare state researcher literature (Scruggs and Allan, 2006; Korpi and Palme, 2004; Korpi and Palme, 1998; Esping-Andersen, 2009; Esping-Andersen, 1991). Hence, even though the World Bank s efficiency measure offers a ranking of countries, it does not offer a blue print for good governance, rather it use example of success to show what success is /../ like telling developing countries that the way to develop is to be developed (Andrews, 2008 p. 383). But even if we imagined a situation where governance in all rich countries looked the same it is not given that rich-country- 5

governance offers a viable governance solution to poor countries simply because rich and poor countries faces very different challenges. The idea of transplanting a governance model into another society is also problematic because it has to be accommodated within a unique context and, not least important, because models of good governance often include contradictory measures. As concluded by Andrews (2008) the idea about a single good governance model is inconsistent, incorrect, unreplicable. But again, if there indeed are different QoG models, we need empirical indicators that are capable to detect these models. Measurement of QoG Commonly used indicators of QoG are based on experts perceptions. Expert valuations suffer from two problems. First, perceptions are likely to be biased, for example, economic growth can be seen as being incompatible with inefficient public administration leading to the perception that countries with growing economies are better administrated regardless of if that is the case or not. Second, the correlation between experts perception of corruption and peoples actual experience of corruption is surprisingly weak, which partly can be explained by biased estimates but also by the diversity of corrupt practices (Donchev and Ujhelyi, 2014; Lin and Yu, 2014). Apart from the causality problem, i.e., if QoG generates economic growth or vice versus, Dellepiane- Avellaneda (2010) points out a set of conceptual problems. Firstly, available QoG measures often conflate institutional constraints and policy choices. New institutionalists distinguish the structure (the institutions specifying the set of players, the permissible moves, the informational conditions, etc.) from the structure-induced equilibrium (the equilibrium in outcomes induced by a given institutional configuration). The problem is that concepts such as political stability and government repression are not institutions-as-constraints, but structure-induced equilibria or institutional outcomes. Using Northean language, they are not referring to the rules of the game, but to the way in which the game is played by certain agents in different arenas (p 205-6). This is apparently challenging. If we in a first instance believe that different political systems, for example democracy compared to dictatorship, produce different policies new institutionalism argues that the outcome from policies differs between countries because the implementation is filtered 6

through different more or less functional institutional systems. But, measures of QoG are usually measuring outcomes such as existing legislations, corruption, transparency, and rule of law. Hence, they are not measuring institutional properties but institutional outcomes. For all practical purpose this is problematic since it both affect measurement validity and the ability to draw the correct policy conclusions, that is, how governments should balance between building good institutions and delivering sound policies. Focusing on a description of the outcome and not the process that lead to the outcome severely limits the explanatory value of the exercise (e.g. Upper, 1974). There can be at least two reasons to why indicators that supposedly are measuring different aspects of QoG in fact measures the same empirical phenomena. First, most countries are following a coherent path of development, simultaneously building the institutions necessary for good governance. That is, only in exceptional cases we find countries that have a developed rule of law but failing democratic institutions or an effective public management but no transparency and accountability. If this is the case the theoretical choice of indicator does not really matter since they are all measuring the same thing, i.e., general QoG. In this case research should focus more on identifying deviant cases and understand particularities rather than the general pattern. Comparative welfare research concerning rich countries is an example this tradition. Traditional QoG measures used when analysing developing countries are more or less useless in this case because among rich countries they almost not discriminate at all. It also follows that comparative cross-sectional analysis that tries to determine whether democracy, rule of law, or effective public management are most important explaining, for example, country differences in educational outcomes or poverty rates are erroneous. What these studies primarily shows is which indicator of QoG that is most reliable and valid. Differences in indicators explanatory power should merely be seen as measurement errors, not guiding lines for policy making. That leads to the second possible explanation, that existing indicators of QoG are not good enough, not able to distinguish between institutional properties and policy outcomes or not enough fine graded to capture important nuances. The large scale Varieties of Democracy project is an example of an attempt to address the latter aspect, the aim being on global scale constructing new and better indicators of democracy and democratization, which falls back on the assumptions that today s inconclusive results regarding the effect of democratization on peoples living conditions are caused by bad measures (REF). 7

Dimensions of QoG In a review of the use of good governance in relation to development policy Gisselquist (2012) points to the elusiveness of the concept and that it reflect a variety of good things that do not necessarily go together in any meaningful way (p 21). Pointing the way forward she distinguish five substantively different aspect of QoG. - Democracy and representation - Human rights: - Rule of law - Government efficiency - Transparency and accountability This is not necessarily the definite categorization of QoG components but it comprises the most central elements and it also tap into to the way different producers of QoG measures more or less explicitly categorize indicators. Once we agree about the central elements of QoG we need to understand the internal relationships, that is, if and in that case how they are causally related. We also need to know what aspects of QoG that possibly is essential for economic growth, positive development of living conditions and other desirable outcomes. We will argue that there is very little explicit disagreement on this point. Our concern is whether it is, given available indicators of QoG, at all possible to do so. Examples of often used existing indicators Below we briefly discuss some of the most commonly used governance indicators. The aim is not to provide a complete and comprehensive overview; we only want to point at some of the common features and problems. Freedom House, a private non-profit organization provides a broad set of indicators rating different aspects of countries governance. Ratings are based on the evaluation of in-house experts subjective perceptions (Arndt and Oman, 2006). Freedom House s measure of Civil Liberties (fh_cl) comprises 15 questions in four categories; freedom of expression, associational and organizational rights, rule of 8

law, and individual rights. Thus, in relation to the theoretical model above the measure is assumed to measure human rights but it also incorporate rule of law. The additional and also frequently used Political Rights (fh_pr) encompass information about electorate process, political pluralism and functioning of government including assessment of corruption. Both measure clearly mix institutional properties, for example electoral processes, and outcomes such as corruption. In relation to the theoretical model the measure covers not only human rights but also rule of law and government efficiency. In some cases, these two Freedom House measures is combined with the Polity IV project s Revised Combined Polity Score (p_polity2), which in turn is calculated by subtracting a measure of Institutionalized Autocracy from Institutionalized Democracy both compilations of yet another subset of indicators (Hadenius and Teorell, 2005; Wahman et al., 2013). The rational for doing so is that the combined measure performs better both in terms of validity and reliability (Teorell et al., 2015). We are inclined to believe that in this case validity and reliability is confused with increased statistical explanatory power. Transparency International since 1995 provides an annual and much used Corruption Perception Index (ti_cpi). The index, obviously tapping into the transparency and accountability box, is described as an surveys of surveys compiling the perceptions of resident and non-resident business people and experts (Arndt and Oman, 2006). The index do not distinguish between administrative and political corruption (Teorell et al., 2015) and is a typical outcome measure and susceptible to both perception biases, i.e., that experts tend to judge corruption in the light of other factors such as economic growth and that experts perceptions not necessarily match with ordinary peoples experiences (Donchev and Ujhelyi, 2014; Lin and Yu, 2014). The World Bank Institute (WBI) provides what has been described as the most comprehensive and probably most carefully constructed publicly available governance indicators (Arndt and Oman, 2006). The WBI indicators covers six areas of governance: 1) Voice and Accountability (wbgi_vae), 2) Political Stability (wbgi_pse), 3) Government Effectiveness (wbgi_gee), 4) Regulatory Quality (wbgi_rqe), 5) Rule of Law (wbgi_rle), and 6) Control of Corruption (wbgi_cce). Also the WBI measures are composite indicators that bring together hundreds of perception indicators from a variety of sources including information from Freedom House. The measures are also mixing incremental 9

institutional properties and policy outcomes. Control of Corruption, Regulatory Quality are clearly perceptions of outcomes, i.e., equilibrium in outcomes induced by a given institutional configuration (Dellepiane-Avellaneda, 2010). Government Effectiveness, Political Stability, and Rule of Law are more ambiguous mixing properties with outcomes while Voice and Accountability possibly is more balanced towards institutional properties. This classification is of course tentative, which mainly reflect the difficulties to se trough all the constituent parts of the measure and the process leading to the final constructs. However, any classification becomes empirically meaningful if the measures actually measures different latent constructs. As shown by Langbein and Knack (2010) there are good reason to doubt that they do. Rather it seems as if they capture one single latent dimension, i.e., the six WBI measures are empirically inseparable. The CIRI Human Data Rights Project provides information about a broad range of human rights in nearly all countries (Cingranelli and Richards, 2010). Data are widely used and information from CIRI is for example incorporated in three of WBI measures (Voice and Accountability, Rule of Law, and Political Stability). Data are, with some minor exemption, based on the annual US State Department Country Reports on Human Rights Practices and covers the same areas, that is: respect of the integrity of the person, respect for civil liberties, respect for political rights, discrimination, and workers rights. As opposed to the previously discussed measures CIRI provides standard based measure, that is, countries are not ranked relative to each other but in relation to standards set in international laws. This is not a trivial difference since it means that it is meaningful to compare single countries development over time, which is not the case when using the relative measures provided by for example the WBI that only give information about relative ranking, not absolute changes. Relating to our theoretical model most of the CIRI indicators relate to human rights but they also tap into other areas such as rule of law and transparency and accountability. The question is: do all these measure aimed at measuring different aspects of QoG in fact measuring separable theoretical aspect of governance or are they all measuring the same thing, i.e., general development? 10

Data and methods Below we describe the indicators used in our analyses. We have ordered them in accordance with the theoretical model suggested by Gisselquist. We are fully aware that the classification of single items can be questioned, but as we will show such discussion might be of theoretical value but is hardly empirically meaningful. Government efficiency: - Government effectiveness (WBI) combines into a single grouping responses on the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies. The main focus of this index is on _inputs_ required for the government to be able to produce and implement good policies and deliver public goods. - Regulatory quality (WBI) includes measures of the incidence of market unfriendly policies such as price controls or inadequate bank supervision, as well as perceptions of the burdens imposed by excessive regulation in areas such as foreign trade and business development. - Functioning of government (EIU) measures to which control over government is exercised by elected representatives, the capability of the civil service, and the pervasiveness of corruption. - Indicator of Quality of Government (IRCG) is an assessment of corruption within the political system. Such corruption is a threat to foreign investment for several reasons: it distorts the economic and financial environment; it reduces the efficiency of government and business by enabling people to assume positions of power through patronage rather than ability; and, last but not least, it introduces an inherent instability into the political process. Democracy and representation: - Political Stability (WBI) combines several indicators which measure perceptions of the likelihood that the government in power will be destabilized or overthrown by possibly unconstitutional and/or violent means, including domestic violence and terrorism - Political rights (FH) measures peoples ability to participate freely in the political process, including the right to vote freely for distinct alternatives in legitimate elections, compete for public office, join political parties and organizations, and elect representatives who have a decisive impact on public policies and are accountable to the electorate. - Revised Combined Polity Score (Polity IV) provide a scale that rank countries from being strongly democratic to strongly autocratic. - Electoral self-determination (CIRI) indicates to what extent citizens enjoy freedom of political choice and the legal right and ability in practice to change the laws and officials that govern them through free and fair elections. - Legislative Electoral Competitiveness (DPI) provides a Scale: 1. No legislature. 2. Unelected legislature. 3. Elected, 1 candidate. 4. 1 party, multiple candidates. 5. Multiple parties are legal but only one party won seats. 6. Multiple parties did win seats but the largest party received 11

more than 75% of the seats. 7. Largest party got less than 75%. - Chief executive a military officer (DPI), yes or no - Finite term in office (DPI) register if there is a constitutional limit of years the executive can serve before election must be called. - Vote fraud (DPI) captures extra-constitutional irregularities. - Municipal government (DPI) measures if municipal governments are locally elected. - Vote share of the largest opposition party (DPI) Human rights: - Civil liberties (FH) measure if civil liberties allow for the freedoms of expression and belief, associational and organizational rights, rule of law, and personal autonomy without interference from the state. - Freedom of assembly and association (CIRI) indicates the extent to which the freedoms of assembly and association are subject to actual governmental limitations or restrictions (as opposed to strictly legal protections). - Political imprisonment (CIRI) indicate the incarceration of people by government o_cials because of: their speech; their non-violent opposition to government policies or leaders; their religious beliefs; their non-violent religious practices including proselytizing; or their membership in a group, including an ethnic or racial group. - Freedom of religion (CIRI) indicates the extent to which the freedom of citizens to exercise and practice their religious beliefs is subject to actual government restrictions. - Freedom of speech (CIRI) indicates the extent to which freedoms of speech and press are affected by government censorship, including ownership of media outlets. - Physical integrity rights (CIRI) measures the existence of torture, extrajudicial killing, political imprisonment, and disappearance. - Women s political rights (CIRI) include a number of internationally recognized rights: The right to vote, the right to run for political office, the right to hold elected and appointed government positions, the right to join political parties, the right to petition government officials. Rule of law: - Rule of law (WBI) includes several indicators which measure the extent to which agents have confidence in and abide by the rules of society. These include perceptions of the incidence of crime, the effectiveness and predictability of the judiciary, and the enforceability of contracts. Together, these indicators measure the success of a society in developing an environment in which fair and predictable rules form the basis for economic and social interactions and the extent to which property rights are protected. - Independency of the judiciary (CIRI) indicates the extent to which the judiciary is independent of control from other sources, such as another branch of the government or the military. - Legal structure and security of property rights (FI) is an index where a low value indicate no judicial independence, no trusted legal, no protection of intellectual property, military interference in rule of law, and no integrity of the legal system. 12

Transparency and accountability: - Voice and accountability (WBI) includes a number of indicators measuring various aspects of the political process, civil liberties and political rights. These indicators measure the extent to which citizens of a country are able to participate in the selection of governments. This category also includes indicators measuring the independence of the media. - Control of corruption (WBI) measures perceptions of corruption, conventionally defined as the exercise of public power for private gain. The particular aspect of corruption measured by the various sources differs somewhat, ranging from the frequency of additional payments to get things done, to the effects of corruption on the business environment, to measuring grand corruption in the political arena or in the tendency of elite forms to engage in state capture. - Economic influences over media content (FH) examine the economic environment for the media. This includes the structure of media ownership; transparency and concentration of ownership; the costs of establishing media as well as of production and distribution; the selective withholding of advertising or subsidies by the state or other actors; the impact of corruption and bribery on content; and the extent to which the economic situation in a country impacts the development of the media. - Political pressures and controls on media content (FH) evaluate the degree of political control over the content of news media. Issues examined include the editorial independence of both state-owned and privately owned media; access to information and sources; official censorship and self-censorship; the vibrancy of the media; the ability of both foreign and local reporters to cover the news freely and without harassment; and the intimidation of journalists by the state or other actors, including arbitrary detention and imprisonment, violent assaults, and other threats. - Laws and regulations that influence media content (FH) encompasses an examination of both the laws and regulations that could influence media content and the government s inclination to use these laws and legal institutions to restrict the media s ability to operate. Freedom House assesses the positive impact of legal and constitutional guarantees for freedom of expression; the potentially negative aspects of security legislation, the penal code, and other criminal statutes; penalties for libel and defamation; the existence of and ability to use freedom of information legislation; the independence of the judiciary and of official media regulatory bodies; registration requirements for both media outlets and journalists; and the ability of journalists groups to operate freely. - Repressive actions (FH) reflect actual press-freedom violations (killing of journalists, physical violence against journalists or facilities, censorship, self-censorship, harassment, expulsions, etc). - Corruption perceptions (TI) focus on corruption in the public sector and define corruption as the abuse of public office for private gain. 13

Methods Confirmative factor analyses using a structural equation modelling (SEM) framework. Models estimated with Mplus. All measures are adjusted so that a higher value always indicates better QoG, more democratic and so on. We use a selection of countries (n=128, see appendix). A number of very small nations and nations with a high degree of missing data are excluded. Analysis Figure 1 shows a confirmative factor model or, to follow the SEM terminology, measurement model that includes the five different QoG elements pointed out by Gisselquist. These five elements make up the first level of latent variables. With almost no exceptions in all cases the correlations between the manifest and latent variables are very high. Thus, if we only look at the relation between the manifest variables and the first level latent variables we could easily conclude that we have very good measures of all five QoG dimensions. But, this is not really the case. The second level latent variable QoG brings together the common variance in the five first level latent variables. As can be seen the correlations between the second level latent variable QoG and the first level latent variables are extremely high, suggesting that there only exists one underlying latent construct. To underline this result we can compare the fit of the model in Figure 1, which, despite the strong loadings, is not very good (RMSE=0.074, TLI=0.610, and CFI=0.640), with a model that assume zero correlations between the first level latent variables, i.e., if we assume that they are truly unique measures. The latter model does represent data significantly worse (RMSE=118, TLI=0.000, and CFI=0.066). The results from Figure 1 indicates that current measures of QoG are incapable of providing distinct measures of the five dimensions suggested by Gisselquist or, for that matter, any other distinct theoretical construct they all more or less, most often more, measure the same latent phenomena. We will return to the issue what latent phenomenon might be. At the same time, the fact that the SEM model in Figure 1 do not fit data particularly well indicate that error terms, that is manifest variables unexplained variance, correlates. In order to sort out the fact that we have one strong common factor and correlated residuals we will in the next step estimate a nested model with two latent factors. 14

In the nested model, Figure 2, we have removed all the first level latent variables from Figure 1 and let the latent overall QoG variable correlate directly with all latent variables. We do so because Figure 1 shows that they essentially are measuring the same latent phenomenon. However, even though most variables correlates strongly with the latent QoG factor there are manifest variables that correlate less strongly. What it mean is that error terms, the variance in the manifest variables not explained by the latent QoG variable, in some cases are substantial. It is a commonly assumed that error terms ought to be uncorrelated with each other. In reality this is often not the case and here it is definitely not the case. However, we do not consider this to be a problem but as an asset. The SEM environment allows us to model the error terms and extract a residual latent variable that captures the covariation between error terms once we have taken account for the common QoG variance. Hence, the analysis shown in Figure 2 extracts two latent variables: the latent QoG and a latent residual variable. The fit of the model is very good (RMSE=0.033, TLI=0.921, and CFI=0.929). Because the residual latent variable summarize the variance not related to the latent QoG variable the correlation between QoG and the residual latent variable is close to zero, i.e., they are measuring to unrelated phenomena. In fact, the good fit of the model presented in figure 2 is calculated under the assumption that the correlation is fixed to zero. But the question remains, what do we measure? Table 2 list all the indicators and their correlations with the latent constructs. Looking firs at QoG we can see that it correlates almost one to one with the WBI indicators, Transparency International s corruption measure and some other well-known measures. The correlations are less strong when looking at measures that more explicitly are designed to measure democratization and liberal human rights. These indicators are on the other strongly related to the residual factor starting with POLITY IV s measure of the degree of democracy, followed by freedom of religion, freedom of assembly and associations, political rights et cetera. Thus, as it seems the residual latent factor is measuring democracy and human rights. Henceforth we will refer to the residual factor as the residual democracy measure. The important thing to be aware is that it does so net of QoG, that is, if a country score high on the residual factor it has more of democracy and human rights than we would expect given their score on the latent QoG variable. But what is actually the latent QoG variable measuring? 15

In Figure 3 the relationship between the latent QoG construct and GDP per capita is shown. An ocular look reveals that there is a strong relationship in lower left corner we find the poor nations and in the upper right corner all the rich countries are found. The correlation is.79. However, there are some deviant cases, countries that do not have as high QoG as their GDP per capita would suggest. One of them is Singapore, which is no surprise since Singapore regularly shows up as an outliner in analysis of governance. The other outliners are exclusively countries dominated by oil and gas industry. Here we find Saudi Arabia, the Gulf States and Norway. If we eliminate the outliners from the analysis the correlation between the QoG measure and GDP per capita increases to.87. Hence, QoG is mainly measuring economic development, which also lead to the conclusion that the residual latent democracy is measuring democratization and human rights net of economic development. Figure 4 shows the relationships between the latent democracy measure and GDP per capita. As expected the correlation is weak. It is also negative but that is referable to the outliners in the upper left corner, i.e., the middle east oil states, i.e., a number of rich states that lack democratic and human rights. Abolishing these countries will lead to a correlation with GDP per capita that is close to zero and that is just as it should be because the correlation between QoG and the residual democracy is zero and QoG is basically just another way of measuring GDP per capita. Looking at Figure 4 we see that the rich, predominantly western, countries are found in the middle of the distribution scoring about zero on the democracy measure. That means that they have as much democracy as we would expect given their QoG score. Countries like China and Cuba on the other hand have less democracy than expected given their QoG score, while for example Poland, Dominican Republic, and Uruguay score higher on the democracy factor than expected given their QoG score. Just as an illustration we have in a final analysis included yet another latent variable measuring the existence of social rights, that is, right to paid sick leave, maternity leave, unemployment benefit, and family benefit. The ILO provides the manifest indicators. Table 2 shows the focal correlations from the additional SEM analysis. The first part of the table includes all countries while the second part excludes the outliners discussed above. Both models shows that once we take account for QoG and GDP per capita, democracy does not affect the occurrence of social rights. Including the outliners lead to a model that shows a significant impact of QoG on social rights but not of GDP. Excluding outliners 16

lead to a model where both QoG and GDP per capita have a positive impact on social rights. In both model, in particular the latter, the correlation between QoG and GDP per capita is very high, indicating that they are not separable and that any partitioning between them are more or less haphazard, which is clearly shown by the effect of excluding a few outliners. Conclusions (very preliminary) Our results shows that widely used indicators of governance are incapable of distinguish between different aspect governance or, following the terminology applied in this paper, quality of government (QoG). They are tapping into one single factor and that one single factor is basically an approximation of GDP per capita. From that follows that most studies that analyses the relative contribution of these indicators trying to say, for example, if rule of law are more important than voice and accountability, when it comes to explain country variation in health outcomes are flawed. Analysis that tries to determine whether governance or GDP per capita are also flawed as most indicators of governance in practice are alternative indicators of GDP per capita. We can also conclude that even though a theoretical model of different aspects of QoG is much desirable, existing indicators are incapable of distinguish distinct aspects of governance. Using a SEM framework we were able extract a residual factor measuring democratization and liberal human rights net of QoG, that is if countries are more or less democratic than what is to expected given their QoG, keepin in mind that QoG more or less equals GDP per capita. Our preliminary results indicates that the contribution of this factor explaining variance in social rights is nil, hence we were able to confirm earlier results indicating democratization per see do not impact if or if not people have guaranteed social rights. The residual factor is based on the fact that error terms among a subset of indicators are correlated. Based on that knowledge a possible tempting way forward would be to lift out these variables from the measure of QoG and create a straightforward democracy index and at the same time letting the remaining indicators measuring QoG. However, that is not a viable solution. Such a procedure will again blur the measures creating two indexes that correlate strongly (0.73), i.e., measuring more or 17

less the same thing. This will happen because the variance belonging to the QoG factor again will be included in the Democracy factor. What to do? The analysis presented here is based on cross-sectional data only. There is a possibility that QoG indicators are ordered in accordance to dominating time sequencing, for example, that rule of law usually precede other QoG features and also economic growth. However, we doubt that is the case. If it indeed would have been the case we should have been able to distinguish several of residual factors as sets of countries would have combined scores on the QoG indicators in distinct ways. We only find one residual factor, but that at least give the opportunity to further analyse the relationship between QoG and democratization over time. A possible more sever problem is the close relationship between QoG and GDP per capita. Again longitudinal data might show a causal relationship between QoG and GDP per capita but again we doubt, because the only outliners we find, except Singapore are countries with large oil production. Otherwise we do not really find countries that have high QoG but low GDP per capita or vice versus. It might be that we need better measures of QoG, measures not affected by the biases discussed above. When it comes to measures of democracy, this is already happening within the variety of democracies project. But it might also be that aspects of QoG and development of QoG are so entwined that a systematic pattern cannot be found. We think that this is highly possible. This leads to the conclusion that future research need to be more focused on what government actually do when it comes legislations and policies. We also need to concentrate more on the interaction between actual policies and other aspects, such as democratization, GDP per capita and, if possible to measure, for example corruption. We also think that a way forward is to compare countries with similar policies but different outcomes. What we clearly should not do is to, for example, compare the impact of different WBI measures and draw conclusions that control of corruption is more important than political stability. What we certainly not should do is to use cross sectional data and let one governance indicator predict another. to be continued. 18

Figure 1. Higher order measurement model 19

Figure 2. Nested measurement model 20

Table 1. Correlations between latent variables and manifest indicators extracted from the nested model Latent QoG variable Latent residual (democracy) variable Regulatory quality (WBI) 0.99 Combined polity score (Polity IV) 0,78 Government effectiveness (WBI) 0.98 Freedom of religion (CIRI) 0,72 Rule of Law (WBI) 0.96 Freedom of assembly and association (CIRI) 0,67 Corruption perception (TI) 0.92 Political rights (FH) 0,64 Control of corruption (WBI) 0.90 Electoral competitiveness (DPI) 0,61 Voice and accountability (WBI) 0.88 Vote share of the largest opposition party (DPI) 0,60 Political stability (WBI) 0.87 Legislative and regulatory media control (FH) 0,59 Legal structure and property right security (FI) 0.86 Civil liberties (FH) 0,57 Indicator of quality of government (IRCG) 0.83 Finite term in office (DPI) 0,56 Functioning of government (EIU) 0.81 Electoral self-determination (CIRI) 0,49 Civil liberties (FH) 0.75 Economic influence over media (FH) 0,47 Physical integrity rights (CIRI) 0.74 Voice and accountability (WBI) 0,46 Economic influence over media (FH) 0.73 Women s political rights (CIRI) 0,45 Independency of judiciary (CIRI) 0.72 Freedom of speech (CIRI) 0,44 Vote fraud (DPI) 0.68 Political imprisonment (CIRI) 0,44 Political rights (FH) 0.67 Municipal government (DPI) 0,40 Legislative and regulatory media control (FH) 0.67 Political control over media content (FH) 0,40 Freedom of speech (CIRI) 0.63 Functioning of government (EIU) 0,38 Repressive action in relation to media (FH) 0.57 Independency of the judiciary (CIRI) 0,30 Electoral self-determination (CIRI) 0.54 Chief executive a military officer (DPI) 0,29 Chief executive a military officer (DPI) 0.52 Combined political score (Polity IV) 0.47 Political imprisonment (CIRI) 0.46 Freedom of assembly and association (CIRI) 0.43 Finite term in office (DPI) 0.42 Freedom of religion (CIRI) 0.41 Municipal government (DPI) 0.37 Women s political rights (CIRI) 0.34 Vote share of largest opposition party (DPI) 0.27 Electoral competitiveness (DPI) 0.23 21

Figure 3. Relationship between QoG and GDP per capita (Correlation.79, excluding outliners.87) 22

Figure 4. Relationship between residual democracy measure and GDP per capita 23

Table 2. SEM* of the relationship between Social rights, QoG, Democracy, and GDP per capita All Social rights QoG Democracy QoG 0.58 Democracy (0.19) - GDP per capita (-0.12) 0.73 - Outliners excluded Social rights QoG Democracy QoG 0.31 Democracy (0.06) - GDP per capita 0.38 0.88 - * The model is an extension of the model presented in figure 2. Apart from governance variables are also five manifest indicators measuring the existence of social rights included. The social rights indicators have been used to extract a latent social rights variable. Only the estimated correlations between latent variables are displayed. 24

References Agnafors, M. (2013) 'Quality of Government: Toward a More Complex Definition', American Political Science Review, 107 (3): 433-445. Andrews, M. (2008) 'The Good Governance Agenda: Beyond Indicators without Theory', Oxford Development Studies, 36 (4): 379-407. Arndt, C. & Oman, C. (2006). Uses and Abuses of Governance Indicators. In: Oecd (ed.) Development Centre Studies. Paris. Banerjee, A. V. & Duflo, E. (2011) Poor Economics - a Radical Rethinking of the Way to Fight Global Poverty New York, PublicAffairs. Cingranelli, D. L. & Richards, D. L. (2010) 'The Cingranelli and Richards (Ciri) Human Rights Data Project', Human Rights Quarterly, 32 (2): 401-424. Daoud, A., Halleröd, B. & Sapir, D. G. (2014) 'Eoccurring Natural Disasters, Quality of Government, and Severe Child Deprivation a Comparative Analysis of 67 Developing Countries', Plos One, Submitted. Dellepiane-Avellaneda, S. (2010) 'Review Article: Good Governance, Institutions and Economic Development: Beyond the Conventional Wisdom', British Journal of Political Science, 40 195-224. Donchev, D. & Ujhelyi, G. (2014) 'What Do Corruption Indices Measure?', Economics & Politics, 26 (2): 309-331. Esping-Andersen, G. (1991) The Three Worlds of Welfare Capitalism, Oxford, Polity Press. Esping-Andersen, G. (2009) The Incomplete Revolution: Adapting Welfare States to Women's New Roles Polity Press. Gisselquist, R. M. (2012). Good Governance as a Concept, and Why This Matter for Development Policy. Helsinki: UNU-WIDER. Glaeser, E. L., La Porta, R., Lopez-De-Silanes, F. & Shleifer, A. (2004) 'Do Institutions Cause Growth?', Journal of Economic Growth, 9 (3): 271-303. Hadenius, A. & Teorell, J. (2005) 'Cultural and Economic Prerequisites of Democracy: Reassessing Recent Evidence', Studies in Comparative International Development, 39 (4): 87-106. Halleröd, B., Rothstein, B., Adel, D. & Nandy, S. (2013) 'Bad Governance and Poor Children: A Comparative Analysis of Government Efficiency and Severe Child Deprivation in 68 Low- and Middle-Income Countries', World Development, 48 19-31. Holmberg, S. & Rothstein, B. (eds.) (2012) Good Government: The Relevance of Political Science, Cheltenham: Edward Elgar Publishing. Korpi, W. & Palme, J. (1998) 'The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries', American Sociological Review, 63 (5): 6661-687. Korpi, W. & Palme, J. (2004) 'New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 Countries 1975-1995', American Political Science Review, 97 (3): 425-446. Langbein, L. & Knack, S. (2010) 'The Worldwide Governance Indicators: Six, One, or None?', Journal of Development Studies, 46 (2): 350-370. Lin, M.-W. & Yu, C. (2014) 'Can Corruption Be Measured? Comparing Global Versus Local Perceptions of Corruption in East and Southeast Asia', Journal of Comparative Policy Analysis, 16 (2): 140-157. North, D. C. (1990) Institutions, Institutional Change and Economic Performance, Cambridge, Cambridge University Press. North, D. C., Wallis, J. J. & Weingast, B. R. (2009) Violence and Social Orders: A Conceptual Framework for Interpreting Recorded Human History, Cambridge, Cambridge University Press. Ostrom, E. (1990) Governing the Commons: The Evolution of Institutions for Collective Action, New York, Cambridge University Press. Przeworski, A., Alvarez, M. E., Cheibub, J. A. & Limongi, F. (2000) Democracy and Development: Political Institutions and Well-Being in the World, 1950 1990, Cambridge, Cambridge University Press. 25