Abstract

Abstract The development and viability of the non-governmental organization (NGO) sector varies across the post-communist world. We explore the impact of corruption on NGO sustainability-the overall enabling environment and activities of the NGO sector-in Eastern Europe and the former Soviet Union from 1998 to 2007. To test hypotheses about the relationship between corruption and NGO sustainability, we employ time-series cross-sectional analyses of 27 post-communist states, controlling for domestic factors such as economic development, government expenditure, and democracy, and international factors such as levels of trade, foreign direct investment, and foreign aid, as well as a country's status vis-à-vis the European Union. We conclude that corruption is consistently and strongly associated with lower levels of NGO sustainability. In particular, our analyses suggest that corruption is likely to degrade the legal environment and fiscal viability of the NGO sector greater than other aspects related to NGO activities such as advocacy or organizational capacity. Résumé Dans les pays à l'ère post-communiste, le secteur des organisations non gouvernementales (ONG) connaît une réussite variable en termes de développement et de viabilité. Nous explorons ici les effets de la corruption sur la pérennité des ONG (environnement global favorable et activité du secteur des ONG) en Europe de Voluntas DOI 10.1007/s11266-013-9404-3 l'Est et dans l'ex Union Soviétique entre 1998 et 2007. Pour tester les hypothèses concernant les relations entre la corruption et la pérennité des ONG, nous procédons à des analyses transversales de séries temporelles pour 27 états post-communistes, en observant des facteurs (variables ?) nationaux (domestiques ?) tels que le développement économique, les dépenses gouvernementales et la démocratie, ainsi que des facteurs internationaux tels que les quantités d'échanges, les investissements directs provenant de l'étranger, l'aide internationale et le statut (la position ?) du pays vis-à-vis de l'Union européenne. Nous concluons que la corruption est systématiquement et fortement associée à une faible pérennité des ONG. Nos analyses suggèrent plus particulièrement que la corruption est susceptible de dégrader l'environnement juridique et la viabilité fiscale du secteur des ONG davantage que les autres aspects des activités des ONG tels que la promotion d'une cause ou leur capacité organisationnelle. Zusammenfassung Nicht-staatliche Organisationen in der post-kommunistischen Welt variieren in ihrer Entwicklung und Ü berlebensfähigkeit. Wir untersuchen die Auswirkungen von Korruption auf die Nachhaltigkeit nicht-staatlicher Organisationen, d. h. die günstigen Rahmenbedingungen und Aktivitäten des Sektors insgesamt, in Osteuropa und der ehemaligen Sowjetunion über den Zeitraum von 1998 bis 2007. Um die Hypothesen über das Verhältnis zwischen Korruption und der Nachhaltigkeit nicht-staatlicher Organisationen zu testen, wenden wir ZeitreihenQuerschnitt-Analysen (Time-Series-Cross-Section-Analysen) von 27 ehemals kommunistischen Ländern an und überprüfen inländische Faktoren, wie beispielsweise die wirtschaftliche Entwicklung, staatliche Ausgaben und Demokratie, und internationale Faktoren, z. B. Handelsumfang, ausländische Direktinvestitionen und Auslandshilfe, sowie den Status eines jeden Landes mit Hinblick auf die Europäi-sche Union. Wir kommen zu dem Schluss, dass Korruption durchweg eng mit einer niedrigeren Nachhaltigkeit nicht-staatlicher Organisationen in Verbindung steht. Unsere Analysen weisen insbesondere darauf hin, dass Korruption mit aller Wahrscheinlichkeit die rechtlichen Rahmenbedingungen und die finanzielle Ü berlebensfähigkeit des Sektors mehr schwächt als andere Aspekte im Zusammenhang mit den Aktivitäten nicht-staatlicher Organisationen, wie beispielsweise die Interessenvertretung oder die organisationale Fähigkeit. Resumen El desarrollo y la viabilidad del sector de las organizaciones no gubernamentales (ONG) varían en el mundo post-comunista. Exploramos el impacto de la corrupción sobre la sostenibilidad de las ONG -el entorno propicio global y las actividades del sector de las ONG -en Europa del Este y en la antigua Unión Soviética desde 1998 a 2007. Para probar la hipótesis sobre la relación entre la corrupción y la sostenibilidad de las ONG, empleamos análisis transversales de series temporales de 27 estados post-comunistas, controlando factores internos tales como el desarrollo económico, el gasto gubernamental y la democracia, y factores internacionales, tales como los niveles de comercio, la inversión directa extranjera y la ayuda externa, así como también el estatus del país frente a la Unión Europea. Concluimos que la corrupción está invariable y fuertemente asociada a menores niveles de sostenibilidad de las ONG. En particular, nuestro análisis sugieren que es Voluntas 123 probable que la corrupción degrade el entorno legal y la viabilidad fiscal del sector de las ONG en mayor medida que otros aspectos relacionados con las actividades de las ONG, tales como la defensa o la capacidad organizativa. Keywords NGO sustainability Á Corruption Á Legal environment Á Financial viability Á Eastern Europe and Former Soviet Union Substantial variation exists in the robustness and sustainability of the nongovernmental organization (NGO) sector in the post-communist states of Eastern Europe and the former Soviet Union. Despite shared historical legacies among many countries in the region (Ekiert and Hanson 2003), the development of the NGO sector has been inconsistent. This is true both geographically across Eastern Europe and Eurasia, and temporally, with individual countries showing both improvement and backsliding. We present an argument that levels of societal corruption, a previously ignored factor, should significantly inhibit the development of an environment supportive of NGOs, and test this argument using cross-national data on various aspects of the NGO sector in the post-communist region. Civil society consists of a wide array of non-governmental, non-profit voluntary organizations operating independently of businesses and government, aiming to realize specific goals (Howard 2002; We start from the premise that NGOs operate and exist in an environment similar to other organizations, and contend that to adequately explain variation in the strength of the NGO sector one must take into account the level of corruption found in the environment in which they operate. There are two main reasons for this. First, corruption impedes the development of a favorable legal environment for the NGO sector. A favorable legal environment includes those legal and regulatory conditions that support NGOs by facilitating new entrants, reducing governmental interference in NGO activities, and easing regulations on taxation and procurement (USAID 2010). Second, corruption degrades the financial viability of NGOs, reducing philanthropic donations and self-financing opportunities from local sources and increasing uncertainty around the costs of doing business. This paper proceeds as follows. First, we discuss NGO sustainability and the USAID Index, an exemplary data source on the NGO sector in Eastern Europe and the FSU. Second, we propose two mechanisms through which corruption should be associated with low levels of NGO sustainability, and address the issue of Voluntas 123 measuring corruption. Third, to test hypotheses about the effects of corruption and other covariates on NGO sustainability, we present a number of models that demonstrate a consistent and robust relationship between corruption and NGO sustainability, and follow this with visualization of the substantive implications of our models. We conclude with a discussion of the implications these findings have for understanding the sustainability of the NGO sector in both the post-communist region and beyond. The Development of NGOs NGOs are civic organizations acting as intermediaries between society and the state. They are collectively organized around specific common interests held by society or some subset thereof, and their primary purpose is neither to pursue profit within the market nor to achieve power within the state Despite this widespread recognition of the vital importance NGO plays, the overwhelming majority of research focuses on domestic and international factors driving the development or penetration of international NGOs (Smith and Wiest 2005; Lee 2010). One unfortunate result of this focus on the role of international (typically well-funded, Western) NGOs has been a lack of attention to the environment in which local NGOs operate: comparative studies of the strength and viability of domestic NGOs are, with some exceptions, rare. In one example, Petrova (2007) examines a cross section of 25 countries, looking at national averages of micro-level variables (interpersonal trust, interest in politics, and activity in voluntary associations) and the aggregate of macro variables (international NGO memberships and GDP per capita). 1 1 Unfortunately, due to having an N of only 25, Petrova aggregates the macro-level variables, including the number of international NGOs. That local NGO development may drive state involvement in international NGOs is not considered. Neither is the possibility that similar factors drive domestic and NGO strength. Voluntas 123 NGO Sustainability in the Post-Communist World After the collapse of communism in Eastern Europe and the Soviet Union, NGOs became active in all areas of social, political, and economic life While the strength of the NGO sector in the region is in dispute USAID's NGO Sustainability Index is a rich data source detailing the viability of the NGO sector, unique in that it tracks and compares progress in all of Eastern Europe and the FSU (Moore 2006). It is a comprehensive measure consisting of seven dimensions critical for the NGO sector: legal environment, financial viability, public image, NGO infrastructure, advocacy, organizational capacity, and service provision (USAID 2008). We categorize these seven dimensions into two subsets: environment and activities. The enabling environment for NGO sectors includes the legal rights, financial conditions, public awareness of NGOs, and the prevalence of organizations and networks supportive of NGO activities. The activity dimension consists of the success of NGO advocacy, the organizational capacity skills found in the sector, and the provision of NGO services for their constituents. Country ratings in the USAID NGO Sustainability Index are based on decisions made by country and regional expert groups, including grassroots NGO staff, international donors, media representatives, academic experts, government partners, legal experts, and USAID experts working in the area of civil society development (USAID 2008). Expert groups in each country discuss every index component, highlighting both existing constraints and progress made. Then the Editorial Committee collects and reviews country reports in a comparative perspective. By Voluntas 123 collecting data this way, the Index allows for comparisons of both setbacks and advances in NGO sector development longitudinally and cross-sectionally. The Index aims for comprehensive and consistent measurements of the strength of NGO sectors in post-communist states. The NGO Sustainability Index is the most thoroughly developed conceptually and extensively measured empirically indicator of the NGO sector. Unlike many existing governance indicators, its focus is explicitly regional and sectoral, and it is not a ''mash-up'' index aggregating disparate and tangentially related indicators (Ravallion 2010) but rather one with explicit conceptual bases and measurement practices. As such, the index does not suffer from the problems found in broader indices such as the World Bank Governance Indicators, which aggregate so much seemingly unrelated information that it is unclear what precisely they are or are not measuring Corruption and NGO Sustainability How corruption-the abuse of public office for private gain-affects the functioning of the state and market is the overwhelming focus of research examining the The state of the economy to support NGO self-financing; the extent to which donation and volunteerism are being nurtured in the local culture, the extent to which government procurement and commercial revenue raising opportunities are developed Public image Public awareness and credibility of the role that NGO play in society The availability of intermediary support organizations that provide NGOs with broad access to local NGO support services Activity Advocacy The formation of coalition and networks to communicate with policy makers as well as the broader public Organizational capacity The presence of constituencies that are transparently governed and publicly accountable, capably managed; exhibition of essential organizational skills Service provision The provision of services that consistently meet the needs, priorities, and expectations of their constituents Source USAID Voluntas 123 consequences of the phenomenon. Many, for example, predict that pervasive corruption is associated with lower levels of economic growth The deleterious effects of corruption on states and markets are well known. How corruption should affect the sustainability of civic organizations is inadequately analyzed. In this section of the paper, we attempt to fill the gap in the existing literature on the consequences of corruption by suggesting how it should affect the working of NGOs and impact the various aspects of NGOs captured in the different dimensions of the USAID Sustainability Index. We focus on the relationship between corruption and the NGO sustainability for two reasons: first, compared to other factors such as the democracy and economic development, the nexus between corruption and the strength of NGO sectors has not been systemically examined. This oversight is important because to the extent that corruption is related to both democracy and development, studies not addressing corruption (provided it is in fact a salient factor) fail to adequately estimate the effects of either variables. Second, despite existing studies of the link between corruption and the market or governments, studies of how societal corruption is associated with the third sector are non-existent, despite widespread attention to NGO actors in the post-communist region and elsewhere. While we hypothesize that higher corruption should be associated with lower levels of NGO sustainability, we also expect that the relationship should differ across the dimensions of sustainability identified by USAID. Broadly speaking, we expect increased levels of corruption to be associated with lower levels of those dimensions of sustainability dealing with the environment in which the NGO sector finds itself, and no consistent relationship between corruption and those dimensions dealing with the activity of NGOs in the post-communist region. In other words, the aggregate Voluntas 123 level of NGO sustainability identified by USAID should be negatively associated with the level of corruption in a given system. This relationship should be stronger for components measuring the NGO environment, and weaker or inconsistent in those components measuring performance. This presents the following: Hypothesis 1 Higher levels of corruption should be associated with a less sustainable NGO environment. Legal Environment According to USAID, ''For an NGO sector to be sustainable, the legal and regulatory environment should support the needs of NGOs. '' (USAID 2008, p. 12) To do this, a legal environment needs to ease barriers and costs to entry, and secure the legal basis to fundraise and engage in other revenue-producing activities. In addition, USAID includes the ''degree to which laws and regulations regarding taxation, procurement, access to information and other issues benefit or deter NGOs' effectiveness and viability.'' When measuring the legal environment, USAID excludes broader legal issues like the rule of law, security of property rights, and judicial independence, focusing solely on legal matters explicitly affecting NGOs. Higher levels of corruption should result specifically in a more precarious legal environment for NGOs primarily because corruption typically increases both the scope of activities of regulatory agencies and the latitude with which these agencies can operate. Critically, systemic corruption should affect agencies such as the tax police, the procuracy, and regulatory agencies to a high degree, given the already-high levels of discretion involved in these fields Hypothesis 2 Corruption should be associated with a poorer legal environments for NGOs. Financial Viability Corruption impedes financial viability for NGOs by significantly increasing costs. As noted above, high levels of corruption raise the overall costs of doing business in 3 Georgia is illustrative of this. Despite scoring well on the legal environment dimension, the level of judicial independence in the country is remarkably low, below either Russia or Ukraine (Linzer and Staton 2011). Voluntas 123 an economy and impede economic performance. Therefore, we should expect higher corruption to be negatively associated with the financial viability of NGOs. Second, corruption should not only raise the absolute costs of doing business for NGOs and thus their viability, but it should also affect them through its increase in the uncertainty of those costs (for a discussion focusing on Russia, see Third, favorable tax regulations help the NGO sector overcome economic hardships to organization Finally, corruption should impede both private philanthropy and the ability of the NGO sector to secure government support. In contexts where small, medium, and large businesses are routinely expected to ''volunteer'' payments to obtain access and services that should be freely available, both the ability and the predilection toward actual philanthropy should be lessened. In instances of widespread corruption, government funding is often directed to political patrons and clients, rather than the NGO sector. In addition, when government contracts are captured through corrupt practices, NGO access to this critical aspect of financial viability should be minimal. This suggests the following: Hypothesis 3 Corruption should be associated with worse financial viability for NGOs. The Other Components of the NGO Sustainability Index There are few reasons to expect any strong first-order relationships between corruption and the other five indicators of NGO sustainability. That is, the manner in which corruption impedes the efficacy of the NGO sector in post-communist countries should work primarily through the legal environment and the financial viability of NGOs. An example of this would be through the lack of any direct effect on the level of NGO infrastructure, as this measures the extensiveness of things such as management training programs and technological support centers, as well as the cohesiveness with which different parts of the NGO sector communicate and work Voluntas 123 together to achieve common objectives. While all of these should certainly be negatively affected by a poor legal environment and a lack of resources, it is through these channels that corruption should inhibit the level of infrastructure. Similarly, the public image dimension measures media support and coverage of NGOs, ''as well as the public's knowledge and perception of the sector as a whole,'' which should be minimally affected by corruption (USAID 2008). The second group of component measures of NGO sustainability focus on NGO sector activities, rather than its environment. While we certainly expect NGOs to be less successful in more corrupt countries, the effectiveness of their operations should be a result of the dimensions discussed above, i.e., mediated through their financial viability and the environment in which they operate. As such, there is little reason to expect corruption to strongly or directly affect the effectiveness of NGOs with regards to advocacy, organizational capacity, and service provision, and, therefore, the statistical relationship between corruption and these components of the index should be weak or inconclusive. For example, the questions the organizational capacity component answers revolve around the efficiency and effectiveness of the internal organization of NGOs, typified by whether ''most NGOs incorporate strategic planning techniques in their decision-making process?'' and to what degree there exists ''a clearly defined management structure within NGOs, including a recognized division of responsibilities between the board of directors and staff members?'' Similarly, the advocacy component addresses issues such as whether ''NGOs formed issue-based coalitions and conducted broad-based advocacy campaigns? Have these campaigns been effective at the local and/or national level in increasing awareness or support for various causes?'' Finally, the service provision component captures the degree to which the NGO sector provides a wide variety of services, asking if that when ''NGOs provide goods and services, do they recover any of their costs by charging fees? Do NGOs have knowledge of the market demand-and the ability of distinct constituencies to pay-for those products?'' (USAID 2008) Unlike the legal and financial measures, these activity measures address dimensions of sustainability that should have a marginal relationship with corruption. This suggests the following: Hypothesis 4 Corruption should not be associated with components of public image, infrastructure, advocacy, organizational capacity, or service provision of NGOs. Data and Methods Due to the nature of missingness in our data, employing listwise deletion (where each observation that has a missing value on one or more covariates is not included) results in removing 40 % of the post-communist country-years between 1997 and 2007. In cases where data are missing at random, there would be no bias introduced by doing so. This is because when there are no systematic patterns to the missingness, listwise deletion is in effect sampling from a larger population. If instead data show systematic patterns of missingness on variables correlated with Voluntas 123 dependent or independent variables, listwise deletion introduces bias As such, we employ multiple imputation, which has been shown to produce substantially less bias than listwise deletion Dependent Variables The dependent variables are drawn from the USAID NGO Sustainability Index. Each dimension is measured yearly in each country and assigned a score on a scale ranging from 1 to 7 (measured to tenths of a point, producing an effectively continuous, sixtypoint scale). As produced by the USAID, the scale codes lower values as more sustainable. Per standard practice, we invert the scale to produce more intuitive estimates of the relationship between NGO sustainability and our independent variables. Therefore, after inversion, a value of seven on the index means that an ''NGO sector's sustainability is enhanced significantly by practices/policies in this area.'' On the other end of the scale, one indicates that an ''NGO sector's sustainability is significantly impeded by practices/policies in this area.'' (USAID 2008) To better understand empirically what these different scores mean, We estimate a number of different models, employing in different instances the overall index score or each individual dimension in separate models. 5 We do this to test the overall relationship between corruption and NGO sustainability, the idea that the legal and financial environment should be more heavily effected by corruption than activities, and finally to test each of the five other dimensions noted above. 4 The World Bank stopped measuring aid received by these states post-accession. Another example is the lack of measures provided by the EBRD with regard to the Central Asian cases. 5 In all models independent variables are lagged one year to prevent simultaneity bias. Voluntas 123 Independent Variables The principal explanatory variable is the level of corruption. We employ the corruption rating from Freedom House's Nations in Transit (2010) series, which evaluates the degree of corruption according to the evaluations of both international and domestic country experts from a variety of independent institutions, and is specifically focused on the post-communist states. 6 Combining and standardizing these inputs, Freedom House produces a measure of corruption that falls between 0 and 10. Since we acknowledge the limitations of corruption index (i.e., subjective measurement from expert groups and slow changes over time), we seek to conduct a robustness check with a different measure of corruption. All models presented in the paper were also estimated using the Transparency International Corruption Perception Index, with negligible differences. No critical covariates lost significance, and coefficients changed minimally. Also, prior to multiple imputation there were more missing values for the TI CPI, bolstering the decision to primarily use the Freedom House rating. Voluntas 123 A number of scholars have raised concerns about the use of perception-based measures of corruption, arguing that expert assessments may not have any consistent relationship to facts on the ground and actual experiences with corruption. Empirical evidence supporting this argument, however, is minimal. In a study of eight African countries, Razafindrakoto and Roubaud (2005) find little relationship between expert assessments and household responses to a survey asking about their practical experiences with corruption. In each country experts predicted higher levels of corruption than reported. It is unclear, however, to what degree this study is generalizable. Indeed, the World Bank and Transparency International perception-based measures are highly correlated (between 0.6 and 0.8) with Global Corruption Barometer survey of 64 countries, as well as the UN's Interregional Crime and Justice Research Institute's survey (Treisman 2006). More importantly for our purposes, there is a consistent and strong correlation between the Freedom House measure and the Business Enterprise and Environment Performance Survey (BEEPS) produced by the European Bank for Reconstruction and Development and the World Bank, which is the most extensive experiencebased study of corruption in the post-communist world (for a further discussion see the Appendix). Furthermore, the fact that there is such high correlation between the different perception-based measures is an argument in favor of the contention that they do in fact capture reality to an acceptable degree. An additional argument is that foreign and domestic-based measures are highly related, with experts both within and outside of countries agreeing. As Treisman (2006, p. 8) notes, ''the evaluations of the Economist Intelligence Units experts in 2005 correlated at r = 0.87 with those of domestic and international business executives surveyed by the World Economic Forum.'' A number of other relevant covariates are included in each of the models, beginning with measures of the level of democracy. We operationalize democracy by employing a combination of the Freedom House and Polity indices, which has been shown to outperform either individual measure in terms of both validity and reliability We also control for external and economic influences on NGO sustainability. As less economically developed countries are less able to fund NGOs domestically, post-communist NGOs rely heavily on international support. We include measures of economic development (GNI per capita) and foreign aid (as a percentage of GNI) to control for those cases in which foreign aid may be augmenting the viability of the NGO sector 7 Government expenditures capture the size and role of government in a given country The role of the European Union in the adoption and development of effective institutions in Eastern Europe has received substantial attention Statistical Models To estimate the relationships hypothesized above, we employ two modeling procedures, each capturing a theorized feature of the data-generating process as well as the complexity of the data at hand. Theoretically, the value of NGO sustainability at any time t should be related to the observed values at prior occasions, such as t -1 . The assumption that observations are identically and independently distributed is tenuous in the case of time-series cross-sectional data, and examinations of the data suggest a high degree of autocorrelation for NGO sustainability. Graphs of the autocorrelation and partial autocorrelation functions for NGO sustainability in each cross-sectional unit suggest a stationary time-series with u \ 1, and panel unit root 7 These three variables are taken from the World Bank's World Development Indicators. Voluntas 123 tests support these impressions. 8 To account for this, we first present a time-series cross-sectional model with a first-order autoregressive component. Similarly, as observations in each given country are repeated measurements of the same unit over time, there is reason to suspect observations within units are not independent (Demidenko 2004). To account for this, we present results of hierarchical linear models with a first-order autoregressive process in the errors. These hierarchical models allows for the intercept to be modeled as a random variable with unit-specific (in this case country-specific) variation. 9 Analysis The models in It is also instructive to look at what fails to reach standard levels of significance in the first two models. The idea-common in the literature on NGOs-that there might exist a trade-off between NGO sustainability and government expenditures finds no support. 11 Similarly, there is no discernible relationship between trade and NGO sustainability, providing no confirmation for the idea that openness to trade and the resulting international connections might support the viability of the NGO sector. An interesting finding is the lack of support for the relevance of international factors. Neither foreign aid nor a country's status as an EU candidate or member 8 Panel unit root tests (Im-Pearsan-Shin and Fisher tests) were conducted in Stata version 11, all results rejecting the null of unit root at p \ 0.01. 9 This contrasts with a fixed effects approach. Fixed effects are inappropriate with time-invariant covariates, and remove the cross-sectional components from the model. As this variation is important, and as the t is small relative to the number of units, mixed effects models are preferred. 10 The coefficient estimate for FDI in Model 1 is -0.008, and a change from the first to third quartile of the observed values for FDI inflows is associated with a less than 0.04 decrease in NGO sustainability. 11 Another possibility is that this is dependent on the level of democracy: if government expenditures are a proxy for state intrusion into the economy, then more intrusive, less democratic governments might have worse environments for NGOs. An interaction term between government expenditures and democracy is, however, highly insignificant and in such a model the individual coefficients barely change. Voluntas 123 shows a statistically significant relationship with NGO sustainability. The latter result is important, as it questions the claims of the ''Europeanization'' literature that the EU accession process has had significant positive effects on a variety of domestic phenomena in Eastern Europe. Results of both the models conclude that once other important variables are controlled for, a country's relationship with the EU has no noticeable effect on the viability of its NGO sector, suggesting a closer look at the efforts of the EU to foster civil society in the accession countries. Looking at the AIC-which penalizes additional parameters-of Models 1 and 2 strongly recommends the inclusion of the additional parameter of a random intercept in Model 2. Model 3 is a restricted version of the preferred Model 2, omitting the control variables that demonstrated no consistent relationship with NGO sustainability. The AIC of the model has greatly decreased (lower values are preferred), far beyond the 9.49 difference for four degrees of freedom minimal for a preferred fit under the 0.05 level of statistical significance of the v 2 distribution. 12 Looking at the results of Model 3 as compared to Model 2, we see that the estimate for economic development has regained statistical significance (excluding it produces a far worse fitting model), and that the estimates and errors for democracy and corruption remain effectively unchanged, as does the AR parameter, u. In other words, excluding the additional parameters that were included to control for hypothesized relationships but which showed no consistent relationships with NGO sustainability significantly improves fit as compared to the full model, suggesting the restricted Model 3 with only three covariates is the preferred model. Model 1 is an AR(1) model, and Models 2-5 are linear mixed effects models with an AR(1) correlation structure. Model 3 is a restricted version of Model 2. The dependent variables of Models 4 and 5 are the legal environment and financial viability components of the index, respectively. Each model has an N of 297 12 This decision is also supported by Bayesian model averaging, which suggests the inclusion of corruption and democracy in 100% of models, and whose averaged posterior means are almost indistinguishable from those in Model 3. Voluntas 123 Hypotheses 2 and 3 address the relationship between the legal environment and financial viability components of the index and our variables of interest. Models 4 and 5 in An interesting feature of Model 4 is the lack of significance of the coefficient for the level of economic development. This suggests that while the level of economic development may be critical for overall NGO sustainability and-for the obvious reasons suggested earlier-the level of financial viability, it is unimportant in determining the legal environment in which NGOs operate. In other words, it may very well be the case that though poor countries find it more difficult to maintain viable NGO sectors, they are nonetheless as capable as wealthier countries in establishing legal rules more friendly to NGOs, a plausible and heartening finding. To save space we refrain from reporting the results of models employing the other five components of the index as dependent variables (output is included in the Supplementary Information). Results largely support Hypothesis 4: the effects of corruption on these aspects of NGO sustainability are either less important than the other components or unimportant in four of the other five components. For organizational capacity and advocacy the only statistically significant covariate in is democracy. When infrastructure is analyzed, democracy and corruption are both significant. The estimated effects of corruption, however, are substantially less than that for financial viability or legal environment, and the standard errors around this estimate are much larger. When service provision is analyzed, both foreign aid and EU status are weakly (p \ 0.05) significant and have a small degree of substantive importance. Corruption and democracy are also both significant, but much like with organizational capacity and advocacy the estimate for the former is smaller than that for legal environment and financial viability. In the fifth instance-the analysis of the public image of NGOs-our hypothesis that the relationship should be weaker or non-existent is unsupported: the estimated effect of corruption on public is as large as its estimated effect on the legal environment (democracy is once again the only other covariate that is statistically significant). A potential explanation for this finding is that the effects of corruption on public trust in institutions would ''travel'' to NGOs and they may be tarred with the same lack of trust. In other words, widespread public distrust of state institutions in the post-communist region might likely be heavily affected by the level of corruption, and this distrust might be easily transferred to the third sector. Further study is called for to better understand this finding. Substantive Implications DeBoef and The solid line in each of the plots in 13 Each column of plots shows changes in one covariate across Models 3-5, and the rows of plots show changes for each covariate within that model: plots a-c shows the expected change in the three covariates in Model 3, plots d-f show changes for Model 4 (legal environment as the dependent variable), and plots g-i show changes for Model 5 (financial viability). Plots do not show what an one unit change at each time period predicts, but rather how the effect of an one-time change at t 1 accumulates over time due to the AR(1) coefficient u reported in each model. This addresses DeBouf and Keele's concerns, explicitly modeling the temporal dependence that exists in the data. It is important to stress that this models a situation in which the only time change in a covariate's value occurs at t 1 . In plot (a), corruption decreases by one unit over the mean (of 4.85) at t 1 and remains at this level for the next half decade, as do all other covariates (which remain at their means). This means that The second and third columns of plots in Of course, it is often the case that when a country decreases state corruption or increases the robustness of its democracy or economy it does so continually over multiple years. Continued increases over multiple years should, according to the models presented here, result in increased NGO sustainability not only in the following year, but also over time. In other words, the relationships reported in Conclusion NGOs play a vital role as bridges between civil society and government. Citizens participate in NGOs both to influence public policy as well as-and this especially holds true in developing and transitioning societies-to provide the collective and all-too-often public goods that go unprovided by the state. In this context, the NGO Sustainability Index measures the favorable conditions for and activities of NGOs in the post-communist world. When they began their transitions after the collapse of communism, the countries in the post-communist region shared common characteristics, such as the lack of NGO development and legacies of state socialism. However, as with many other aspects of state and society (Gryzmala-Busse 2007; Frye 2010), much divergence has occurred in the level of NGO sustainability in two decades since the collapse. To date, little research has been conducted on the potential explanations for this, and our paper has four important implications for the literature. First, we provide the first comprehensive quantitative examination of the determinants of NGO sustainability over time in the former communist states of Eastern Europe and Eurasia. Doing so we employ an underused index developed by the United States Agency for International Development, and more importantly we disaggregate the index. On a theoretical level we offer hypotheses that address the differing effects corruption should have on the different aspects of NGO Voluntas 123 sustainability. Empirically, we demonstrate that the relationships we hypothesize obtain, showing both statistical and substantive significance. The second implication regards what matters. Corruption, democracy, and the level of economic development are all consistently related to the level of NGO sustainability broadly, and specifically to the legal environment and financial viability of the NGO sector. We provide further evidence supporting the extant finding that levels of democracy and economic development are related to the robustness of the NGO sector, both globally and more precisely with reference to the post-communist context. More importantly, we introduce a new explanatory factor of the strength of the NGO sector, and show that it is both substantively and statistically significant. As Third, we offer another site of commonality for the literatures on business and NGOs. While much existing research shows that the business environment is significantly impeded by corruption, no work has sought to examine how corruption might affect the non-profit sector. We find that, similar to the business sector, NGO strength is impeded when corruption is pervasive. This finding speaks to the current debates, supporting recent research arguing that NGOs are best viewed as businesslike organizations operating in a competitive policy market (Prakash and Gugerty 2010). Here, NGOs and firms share similarities including their emergence, competition, and maintenance as actors motivated and influenced not only by principled beliefs but also by instrumental responses to incentive structures. One implication of this perspective on NGOs is that they should likely be affected by corruption similarly as businesses, which this study shows. Our final contribution relates to what we find that does not help explain NGO sustainability. With regards to the broader literature on NGOs, we provide some small support for the contention that there is not a trade-off between government expenditures and the vitality of the NGO sector, as it shows no relationship with any of the eight dependent variables employed in our models. Similarly, not a single international factor shows any consistent relationship with NGO sustainability. Neither FDI inflows nor trade as a percentage of GDP produce statistically significant or substantively important coefficients across models. Foreign aid, which has been suggested by a variety of scholars to be an important contributor to the success of NGOs in developing and transitioning societies, also fails to produce a statistically significant or non-zero coefficient. This finding suggests further research into the question of under what conditions foreign aid may or may not support the development of the NGO sector. With specific reference to the literature on the post-communist experience, we show that a state's status vis-à-vis the European Union has no association with the sustainability of its NGO sector. This offers a strong refutation of the popular idea that the EU's influence in the region has been something akin to an unalloyed good. This is not to say that the European Union has had a negative impact on NGOs in the region, but simply that there is no evidence here to support the idea of the EU as an exporter or protector of a robust NGO sector. This finding provides support for the possibility that studies finding a relationship between EU influence and positive Voluntas 123 outcomes are problematic because they fail to account for possible selection bias regarding EU candidacy and the success of the post-communist transition. Our paper raises broader questions regarding the political, legal, and economic conditions under which NGOs as social actors representing civil society operate, and how corruption influences NGO sustainability. We show that the level of corruption in a given country is, along with development and democracy, a key factor explaining variation in NGO development in the post-communist world. If corruption is rampant, NGOs connecting civil society to the state and market operate under substantially worse conditions. The implications of this study suggest that future research exploring the variation in NGO sustainability cross-regionally is called for. For example, examining variation in Asia, Africa, or Latin America would help generalize our findings linking corruption to NGO (un)sustainability. Such an examination would require, however, comparable cross-national data, which to date only exists for Eastern Europe and the former Soviet Union. At a policy level, the recommendation suggested by this paper is simply that the not insubstantial sums of money being spent promoting the effective development of NGO sectors in the post-communist world and elsewhere need to be spent with greater cognizance of the fact that broader societal factors may work to undermine these efforts. A greater focus on effective anti-corruption policies would be a wise investment if the goal is the strengthening of NGOs and civil society. On the other hand, results also suggest that economic development is not a critical precondition for certain types of reform, namely that of the legal system, which has no relationship with societal wealth, suggesting that even in poorer countries the creation of legal rules protecting and fostering the NGO sector can have positive effects. Acknowledgments The authors would like to thank Eunbi Yu, Daniel Berliner, and the anonymous reviewers for their comments on earlier drafts of this paper. Replication data and code can be found on the authors' websites.