Although the positive developmental effects of infrastructure provisioning are well documented, research on the potential role of governance in the improvement of infrastructure performance and individual-level service utilization is lacking. I explore the effect of infrastructure provisioning on individual-level health service utilization, paying close attention to whether governance at different levels shapes people's access to health care. The different geographical levels of infrastructure provisioning, governance, and health service utilization require a multilevel analysis, which I perform using Afrobarometer Round 5 survey data on 34 African countries in a three-stage mixed-effects modeling. Results show that the presence of health infrastructure is crucial for enhancing people's health service utilization. However, people encounter certain problems when receiving services at their local health clinics or hospitals, and these problems are directly linked with governance in the health sector as well as overall governance at the country level. Improvements in people's health service utilization therefore require both better infrastructure provisioning and better governance at different levels, as the former does not guarantee the latter. Development scholars need to widen their focus beyond national-level governance and help policy makers identify at which level state interventions are most needed for removing barriers to development.
A growing literature identifies a big gap between infrastructure provisioning and people's actual health service utilization (Abdulraheem, Olapipo, and Amodu 2012; Balabanova et al. 2004; Hsia et al. 2012; Jaro and Ibrahim 2012; Kasper 2000). Studies find that many developing countries have accessible health clinics but that people receive few services because doctors and health workers are often absent at the facility and people have to make informal payments to receive services (Alcázar and Andrade 2001; Chaudhury and Hammer 2004; Ensor 2004; Garcia-Prado and Chawla 2006). Chaudhury and his colleagues (2006) report that on average 39 percent of doctors and 31 percent of health workers were absent at the service point during random spot checks in Bangladesh, Ecuador, India, Indonesia, Peru, and Uganda. Low pay was one reason, but the study offers evidence that poor working conditions at the facility were the main cause of such absenteeism. Of course, not all developing countries offer the same scenario. In the Indian state of Kerala, for example, people's health service utilization is high because the state “has a health delivery system that actually functions, with doctors at their posts serving clients rather than enjoying their professional perquisites elsewhere” (Evans 1995:236).
Infrastructure performance is thus closely linked with governance issues, as sufficient infrastructure provisioning—physical infrastructure, staff, and supplies—does not guarantee that people receive services. However, the importance of governance in the infrastructure sector for improving people's service utilization is not well researched, although donor communities often express concerns about the weak infrastructure performance of many developing countries (Kessides 2004; World Bank 1994). In this article, I explore the effect of infrastructure provisioning on individual-level health service utilization, paying close attention to whether governance at different levels shapes individual access to health care. The different levels of infrastructure provisioning, governance, and health service utilization require a multilevel analysis. National-level analyses often fail to capture the mixed effects produced at lower levels because of the large within-country variation in state capacity and therefore may under- or overestimate the effect of institutional characteristics on local-level developmental conditions (Sacks and Levi 2010). Moreover, national-level studies commonly use child health as a developmental outcome, whereas people may receive other health services beyond child health from a health infrastructure. Many of these studies also focus on service delivery rather than service utilization. Shifting the focus from service delivery to service utilization helps us explore the impact of infrastructure provisioning on broader developmental outcomes and enables us to identify the barriers that may prevent individuals from enjoying their entitled services. Service utilization concerns whether services actually reach the target population; identifying barriers helps remove those in order to improve people's lives. This is a growing concern in development literature, thanks to Amartya Sen and others (Evans 2010, 2011; Sen 1983, 1999). In this paper, I link this concern with governance, using empirical evidence from Africa, where the issues of infrastructure provisioning, governance, and health service utilization are very pertinent.
INFRASTRUCTURE PROVISIONING, GOVERNANCE, AND SERVICE UTILIZATION: THE LINK
In the 1970s and 1980s, neoliberal development thinkers identified corruption, rent seeking, and bureaucratic ineffectiveness as the main problems of service delivery in African countries (Bates 1981; Bhagwati 1982; Colander 1984; Krueger 1974; Srinivasan 1985; World Bank 1981, 1984). To improve service delivery, they proposed deregulating markets, downsizing public service, and privatizing service infrastructure. As part of neoliberal structural adjustment policies, developing countries in Africa and elsewhere received considerable foreign direct investment commitments for transferring the operational responsibility of infrastructural projects to the private sector, but there was also a sharp decline in donor support for public infrastructure projects at this time (Kirkpatrick, Parker, and Zhang 2006). For many of these countries, limited investment in infrastructure and human resources has over time reduced state capacity to provide basic services to people (Cornia, Jolly, and Stewart 1987; Epping-Jordan et al. 2005). Yet many neoliberal thinkers have argued that any increase in state capacity, such as through tax revenue extraction undertaken at the expense of the private sector, would hamper economic growth and thereby reduce people's service utilization (Friedman and Friedman 1980; Olson 1982). Studies testing these claims in Africa and elsewhere find mixed results. Some studies find that state capacity or state intervention in terms of state size has positive effects on population well-being (Bradshaw and Huang 1991; Bradshaw and Tshandu 1990; Lena and London 1993), others find negative effects (London and Williams 1990; Moon and Dixon 1985; Shen and Williamson 1997, 2001), and still others find no significant effect (Bradshaw et al. 1993; Frey and Field 2000).
Dawson (2010) claims that the mixed findings often result from the fiscal measures most studies use for state capacity or governance. Using the rule of law as a nonfiscal dimension of governance, he finds that governance has a strong, negative and robust effect on child mortality. He does not find a statistically significant effect of state size measured by tax revenue as a share of GDP when the rule of law and other politico-economic variables are controlled for. The fiscal measures of state capacity are directly related to economic growth, which in turn depends on the quality of governance. There is growing evidence that the quality of governance measured in nonfiscal terms of government effectiveness, democratic accountability, rule of law, and control of corruption positively affects economic growth and service utilization outcomes independent of the size or fiscal strength of the state (Kaufmann and Kraay 2002; Rodrik 1999; Rodrik, Subramanian, and Trebbi 2004). As described below, the various dimensions of governance can affect infrastructure performance in the health sector in different ways.
In most developing countries, the public sector is the largest provider of health infrastructure and health care services. Countries with an effective public sector encourage higher levels of human capital accumulation through meritocratic recruitment into the public service (Baum and Lake 2003; Easterly, Ritzen, and Woolcock 2006). An effective bureaucracy increases the productivity of public spending (Rajkumar and Swaroop 2008; Rauch and Evans 2000) and contributes to economic growth by improving credit ratings, attracting more investment, and ensuring better use of foreign aid resources (Asteriou and Price 2005; Burnside and Dollar 2000). An effective government utilizes its infrastructure to increase government revenues and consequently public spending on health and social welfare (Kiser and Sacks 2009).
According to Vu (2007), “a stable, centralized government, a cohesive bureaucracy, and effective coercive institutions” constitute necessary “developmental structures” that help states play their “developmental roles” (p. 28). In the infrastructure sector, state leaders have to negotiate with industrial capitalists and civil societal groups and direct the coordination of public and private resources toward achieving developmental outcomes related to service utilization (Evans 1995; Heller 2013). An effective government can enhance people's service utilization by incentivizing, regulating, and supporting investment in infrastructure development and industrial activity (Evans, Huber, and Stephens 2014).
Mann (1984, 2008) argues that infrastructurally strong modern democratic states enjoy more legitimacy and stability as they enhance people's service utilization by logistically, as opposed to coercively, implementing decisions. Moon and Dixon (1985) also find that democratic states that channel the preferences of mass publics to the level of state policy are more likely to provide for the basic needs of their populations, irrespective of state strength and the ideological orientation of the ruling class. In contrast, authoritarian regimes are able to extract resources but seldom provide sufficient goods and services to people, and when they provide services to the citizenry they do this to gain civil obedience in exchange for their civil and political rights, which in other ways hinders their service utilization (Mann 2014). In extreme cases, authoritarian regimes may create developmental disasters with their high modernist aspirations and central planning for infrastructure development and service delivery (Scott 1998).
However, there is inconclusive evidence as to whether democratic states are better able to enhance people's health service utilization, since some scholars have found a strong negative relationship between democracy and health (Lena and London 1993; Navia and Zweifel 2003; Powell-Jackson et al. 2011), while others have found no significant relationship (Gerring, Thacker, and Alfaro 2012; Halleröd et al. 2013; Ross 2006; Shandra et al. 2004). In a recent study, Wullert and Williamson (2016) find that infant mortality rates are higher in hybrid regimes and lower in both democracies and autocracies and claim that hybrid regimes divert attention and resources away from health care in ways that have detrimental effects on health service utilization. External pressure and conditional foreign aid for democratization may not work, since long-term dependence on foreign aid often reduces governments' need to tax their populations and may reduce public pressure for democratic accountability and service provisioning (Bräutigam 2000; Knack 2000; Moore 1995; Svensson 2000).
In situations like this, a strong rule of law can help prevent government officials from abusing their powers through the impartial application of laws and enforcement of penalties, which can positively affect health service utilization (O'Donnell 2004). Dawson (2010) proposes two other mechanisms through which the rule of law can influence health service utilization in terms of child health. These mechanisms are mainly related to state-society coordination structures, where a strong rule of law can help improve market efficiency and strengthen civil society. Regarding the first mechanism, Dawson argues that a strong rule of law may reduce transaction costs of market exchange by enforcing contracts and property rights. Efficient markets can facilitate economic growth, provide necessary resources to deliver adequate public services, and increase private health care alternatives where public provisioning of health services is inadequate. Regarding the second, Dawson argues that a strong rule of law may strengthen civil society organizations to partner with the state to provide better health services to people.
Closely related to the rule-of-law dimension is a state's capacity to control corruption. In the infrastructure development sector, corruption is more widespread; this decreases the value of investment by foreign and domestic investors and undermines economic growth in developing countries (Alesina and Weder 2002; Transparency International 2011). Corruption in construction lowers the standard of the infrastructure and requires expensive repair and maintenance, and the loss of revenue and diversion of public funds associated with this may cause governments to spend less on infrastructure services (Dal Bó and Rossi 2007). Corruption can divert resources away from social sectors, such as health and education, and can bias resource allocation toward major infrastructure projects because of opportunities for financial kickbacks and political patronage (Kenny 2006; Sohail and Cavill 2008). Corruption also lowers people's health service utilization level by contributing to ghost and absent doctors and health workers, informal payments for services, siphoning off of medical supplies to market, and elite capture of infrastructure services (Chaudhury and Hammer 2004; Chaudhury et al. 2006; Ensor 2004; Lewis 2006, 2007; World Health Organization 2006). However, transportation and communications networks can closely monitor government officials, doctors, and health workers to ensure that they serve in their posts and provide services to people. Controlling corruption also makes possible the improvement of working conditions at the service point and the prevention of emigration of valuable and experienced human resources (Globalisation Knowledge Network 2007).
In summary, the literature suggests that countries with an effective government and effective institutions for democratic accountability, law enforcement, and control of corruption can increase state-society partnership in development and produce better developmental outcomes, including improved levels of health service utilization. As Sacks and Levi (2010) find, people living in countries with an effective government, reliable law enforcement, and improved infrastructure enjoy, on average, higher levels of food security. They claim that there is more variation within than between countries in government infrastructure performance. In the following section, I use multilevel modeling to explore how governance at different levels affects state infrastructure performance for enhancing service utilization at the individual level.
DATA AND METHODS
Data on infrastructure provisioning, governance, and individuals' health service utilization comes from 34 African countries surveyed by Afrobarometer between October 2011 and June 2013 (Afrobarometer 2015). Data collectors recorded if there was a health clinic, a piped water supply, an electricity grid, a sewage system, and a paved road that most houses could access in a primary sampling unit (PSU) or census enumeration area. In most cases, eight households were randomly selected from each PSU for data collection. The sample was further stratified by key social characteristics in the population such as subnational area (e.g., region or province and district) and residential locality (urban and rural).
The hierarchical nature of the Afrobarometer data requires a multilevel modeling strategy. In this study, I reconfigure the Afrobarometer data into a three-level hierarchical structure, with variables used from each of the following three levels: individual respondents (level 1) nested in PSUs (level 2), which in turn are nested in countries (level 3). When individual-level observations are nested into higher-level units, the observations are no longer independent, and hence a pooled estimator cannot be used to correctly estimate the varying intercepts and slopes at each level. Since intercepts may vary across countries, and since the individual-level variables may have unequal slopes across countries, we need to use multilevel mixed-effects models to correct for biases in parameter estimates and standard errors (Gelman and Hill 2007; Snijders and Bosker 1999). Mixed-models estimate both fixed effects as standard regression coefficients and random effects as a sum of variance and covariance parameters. The fixed effects are estimated controlling for all time-invariant differences between individuals such as culture, religion, nationality, ethnicity, and gender, so the estimated coefficients cannot be biased because of omitted time-invariant characteristics.
I also run models excluding infrastructure variables to see how governance alone shapes individuals' health service utilization outcomes. These outcomes are related to specific problems that they encounter while receiving services at their local clinics. Notably, there is wide variation in infrastructure provisioning by rural-urban sampling units. The data of this study are weighted to correct for either deliberate (e.g., to provide an adequate sample of specific subgroups for analytical purposes) or inadvertent over- or undersampling of particular sample strata for some Afrobarometer Round 5 surveys (Afrobarometer 2015). In those cases, Afrobarometer includes weighting factors in the data set, and I use both within-country and across-country weighting factors to account for the variation in sampling when national-level statistics are calculated.
Afrobarometer surveys have a number of variables to measure health service utilization at the individual level. I combine the following three variables using exploratory factor analysis (EFA) to obtain factor scores on my health service utilization dependent variable, with higher scores corresponding to better health service utilization:
Gone without medicine or medical care. Afrobarometer interviewers asked the respondents a crucial question about their health service utilization: “Over the past year, how often, if ever, have you or anyone in your family [involuntarily] gone without medicines or medical treatment?” Responses to this question are coded as 1 = never, 0 = at least once.
Received free visits or medicines. Respondents were asked: “Do you or anyone in this household receive any of the following: Free visits or medicines from a public or government-run health clinic or hospital?” Responses to this question are coded as 1 = yes, 0 = no.
Had difficulty receiving medical treatment. Respondents were asked: “Based on your experience, how easy or difficult is it to obtain the following services from government . . . Medical treatment at a public clinic or hospital?” Responses to this question are coded as 1 = very easy/easy, 0 = difficult/very difficult. Those who never tried to get the service or reported “don't know” were treated as missing values and excluded from the analysis.
I also combine six variables using EFA to create another dependent variable regarding problems encountered with local clinic, with higher scores corresponding to greater extents of the problems encountered. Respondents were asked: “Have you encountered any of these problems with your public clinic or hospital during the past 12 months: Services are too expensive/unable to pay?” The same question was asked separately for each of the following problems: absent doctors, lack of medicines or other supplies, lack of attention or respect from staff, long waiting time, and dirty facilities. Responses related to each of the six problems are coded as 0 = never or only once or twice, 1 = a few times or often encountered the problem. Those who never had to go to a local health clinic in the year prior to the survey were excluded from the analysis.
The independent variable infrastructure provisioning is measured by the physical presence (coded as 1) or absence (coded as 0) of the following service infrastructures at the PSU level that most houses could access: health clinic, electricity grid, piped water, sewage system, and paved road. Access to a health clinic is directly related to health service utilization, while other infrastructures are indirectly related. For example, an electricity grid is important for operating medical equipment and refrigerating certain medicines. Piped water system is important for supplying clean water that prevents waterborne diseases. Sewage system is related to safe waste disposal and public health. Finally, paved roads help transportation of medical supplies to remote clinics; patients can also reach a clinic more easily and quickly if there is a paved road network in the locality.
Afrobarometer (2015) also offers data on the following dimensions of governance:
Rule of law. This variable is measured by respondents' perceptions about how often (a) the president or prime minister ignores the law or constitution of the country, (b) people are treated unequally under law, (c) government officials who commit crime go unpunished, and (d) common people who commit crime go unpunished. Responses to each of these questions are first coded as 1 = never/rarely, 0 = often/always, and then combined using EFA to obtain factor scores, with higher scores corresponding to better rule of law.
Control of corruption. Respondents were asked to give their perceptions about the extent of corruption in their country among (a) government officials, (b) police, and (c) judges and magistrates. Responses to each of these items are first coded as 1 = none/some of them were corrupt, 0 = most/all of them were corrupt, and then are combined using EFA to obtain factor scores, with higher scores corresponding to lesser extents of corruption, which I take for greater control of corruption.
Satisfaction with democratic rule. Respondents were asked about their level of satisfaction with the democratic rule in their country. Responses are coded as 1 = they were fairly/very satisfied, 0 = either their country was not a democracy or they were not at all/not very satisfied with democracy.
Government effectiveness. Respondents were asked to give their perception about how well or badly their current government was (a) managing the economy, (b) fighting corruption, (c) reducing crime, (d) maintaining roads and bridges, and (e) providing electricity supply. Responses to each of these items are first coded as 1 = well, 0 = badly, and then are combined using EFA to obtain factor scores, with higher scores corresponding to better government effectiveness.
The Afrobarometer data set also allows me to control for potential individual-level determinants of health service utilization such as the age of respondents (in years), level of education (0 = no schooling, 1 = informal schooling only, 2 = some primary, 3 = primary completed, 4 = some secondary, 5 = secondary completed, 6 = college/diploma, 7 = some university, 8 = university completed, 9 = postgraduate), and employment status (1 = employed). I also control for the types of residential area one lives in (1 = urban).
At the country level, I control for several variables that potentially link infrastructure provisioning with service utilization. In most cases, these variables have 2012 values, but in a few cases I use 2011 or 2010 values depending on the availability of the data:
GDP per capita. This is a common measure of economic growth that might be correlated with infrastructure provisioning and health service utilization. I use the natural logarithm of GDP per capita measured in 2011 constant international dollars, PPP adjusted, from the World Bank's (2015) World Development Indicators (WDI) database.
Democratization. To see if people's service utilization status is better in more democratic states, I use a country-level measure of democratization from the data set created by Vanhanen (2014). In the data set, the index of democratization is formed by multiplying the political competition and the political participation variables and then dividing the outcome by 100. The competition variable is calculated by subtracting from 100 the percentage of votes won by the largest party, the party which wins most votes, in parliamentary elections or by the party of the successful candidate in presidential elections. The participation variable is calculated as the percentage of the total population who actually voted in the election.
Net official development assistance (ODA) per capita (in current US$). I use the log of net ODA per capita as a control that may affect infrastructure performance, as Knack (2000) and Svensson (2000) argue. These data come from the World Bank's WDI database.
Foreign direct investment (net inflows as percentage of GDP). Foreign direct investment (FDI) is a common measure of a country's investment dependency, which may affect infrastructure performance and health service utilization (Lena and London 1993; London and Williams 1990). I take the FDI data from the WDI database.
Tax revenue (as percentage of GDP). Tax revenue refers to compulsory transfers to the central government for public spending purposes. It is a common measure of state size or state strength (Shen and Williamson 1997). These data come from the WDI database.
Public spending on health (as percentage of GDP). I control for public spending on health as an infrastructure provisioning in the form of capital flows. It should be considered together with infrastructure provisioning in the form of physical stocks such as health clinics. These data come from the WDI database.
Colonial origins. Infrastructure provisioning in most African countries was largely shaped by their colonial origins, as Cooper (2002) and Herbst (2000) claim. To account for this path dependency, I control for the colonial origins of countries, coded as 1 = never colonized by a Western overseas colonial power, 2 = British, 3 = French, 4 = Portuguese, 5 = Belgian, and 6 = former German colony. These data were collected by Hadenius and Teorell (2007) and come from the compiled data set by Teorell et al. (2015).
A multicollinearity test using all the variables shows that only two variables—tax revenue and health expenditure as percentage of GDP—are correlated at 0.627 level with relatively low tolerance (0.455 and 0.497, respectively). I put these two variables separately and then in the same equation but find no significant difference in the results. For the rest of the variables the test shows no significant multicollinearity issues. In addition, I have performed a residual analysis after running each regression equation to identify potential outlying cases (following Loy 2013). A few outliers are detected, but I find no problematic difference in results when excluding outliers and therefore keep them in the final models.
Table 1 gives the percentages of PSUs with easy access to service infrastructure in 34 African countries, along with their colonial origins and sample size as surveyed by Afrobarometer (2015). The data show that on average service infrastructure was quite widespread in African countries, with the sewage system being the main exception. Overall, some African countries had better infrastructure provisioning than others, and within each country urban areas had better infrastructure provisioning than rural areas. On average, 61 percent of the total sample belonged to rural areas.
While infrastructure provisioning was more or less widespread in 34 African countries, access to important services was not. In response to a question on whether the respondent or anyone in his or her family had ever had to go involuntarily without medicine or medical care, 53 percent reported that they had had to go without medicine or medical care at least once over the year prior to the interview. In response to another question on how easy or difficult it was to obtain medical treatment at a public clinic or hospital, 43 percent of those who had experience with this said that it was difficult or very difficult, while 57 percent said it was relatively easy or very easy. Respondents were asked separate questions regarding if they had ever encountered specific problems with their local health clinic or hospital. Of those who had gone to a public clinic or hospital during the 12 months prior to the interview, 37 percent reported absent doctors; 33 percent said the services were too expensive for them; 43 percent reported lack of medicine or other supplies; 42 percent said they encountered lack of attention or respect from the staff; 62 percent reported a long waiting time, and 32 percent reported that the facilities were dirty (see table 2). Of the total observations, only 26 percent reported that they had received free visits or medicines or both from a local health clinic during the previous year.
Regression estimates show mixed results for infrastructure provisioning on individual health service utilization. As shown in table 3, the presence of a health clinic within the accessible reach of a household is significantly associated with an increase in the probability of an individual's health service utilization, with all else equal. The presence of piped water also remains significant and positively associated with the health service utilization outcome across all models, while the presence of an electricity grid, a sewage system, and a paved road is not significant in the full models. Urban residence is not significant in any model. I include an interaction term for health clinic and urban residence in the final models to test for interaction effects. The interaction term is significant and negatively associated with health service utilization (results not shown in table 3), suggesting that although urban areas have more infrastructure provisioning, people's health service utilization is lower in urban areas compared to rural areas.
While service infrastructures have differential effects, all Afrobarometer governance indicators have strong favorable effects on health service utilization across all models. Of all dimensions, government effectiveness has the largest effect on increased health service utilization. Notably, the Vanhanen index of democratization at the country level is not significant in model 3 of table 3. To see if the relationship between democratization and health service utilization is quadratic, as Wullert and Williamson (2016) claim, I include a square term for democratization in model 4. Now, the square term is significant and positive, suggesting that people in both autocracies and democracies may enjoy better health services compared to hybrid regimes. However, while GDP per capita is significant in model 3, it is not in model 4. To see if this is due to an interaction between GDP per capita and democratization, I add an interaction term in model 5 and find that the interaction term has even a larger effect on health service utilization than the square term. The results suggest that economic growth is needed for a democracy to enhance people's health service utilization. Other interaction terms for GDP per capita and different Afrobarometer governance indicators are included in the final model, but none is significant (results not shown in table 3). Similarly, the interactions between health clinic and those governance indicators are insignificant, suggesting that both health infrastructure and governance have independent and direct effects on the health service utilization outcome.
Besides the physical presence of health infrastructure, public health spending has strong favorable effects on health service utilization in table 3. A one-unit increase in the public spending on health (as percentage of GDP) is associated with a 12.8 percent increase in the health service utilization outcome in model 3 (and a 15.4 percent increase in model 5), all else equal. State size in terms of tax revenue collection (as percentage of GDP) is not significant, nor are net ODA per capita and foreign direct investment (as percentage of GDP). Colonial origins of the country are not found to be significantly associated with people's health service utilization and are not shown in table 3. At both national and PSU levels, all random intercepts (given in standard deviation units) are significant, justifying the use of the nested models in this study. At the same time, they show how much influence the time-invariant characteristics at each level have on the outcome.
As already mentioned, Afrobarometer surveys provide information on the problems people encountered with their local health clinics in 34 African countries. Table 4 presents regression estimates of governance and other variables on a combined dependent variable that represents six such problems: absent doctors, lack of medicines and other supplies, lack of attention and respect, expensive services, long waiting time, and dirty facilities.1 Results show that all four Afrobarometer governance indicators are significantly associated with a decrease in the probability of encountering problems with local health clinics. Country-level democratization also has strong negative (favorable) effects on the outcome when it interacts with the levels of economic growth, although GDP per capita alone is not significant in any model. No other fiscal controls are significant as well. Of the individual-level controls, only urban residence is significantly associated with an increase in the probability of encountering problems with local health clinics. Colonial origin dummies were also entered in the final model but were not found significant and therefore are not shown in table 4.
Table 5 presents the results for each of the six specific problems people encountered with their local health clinics. Consistent with previous results, all governance indicators are highly significant and negatively associated with each of the problems related to health service utilization. The country-level democratization and GDP per capita interaction term is also significant and is negatively associated with each of the problems. Public spending on health is negatively associated with three specific problems, namely absent doctors, expensive services, and dirty facilities, but it is not significant for the other three problems. Foreign direct investment is weakly significant and negatively associated with only one problem, expensive services, suggesting that foreign investment in the medical service sector may play a role in reducing costs of medical services. Net ODA per capita and tax revenue (as percentage of GDP) are not significantly associated with any of the six problems. As for the individual-level controls, those with higher levels of education are less likely to report that services are too expensive but are more likely to report absent doctors and dirty facilities. Compared to rural areas, those living in urban areas are more likely to encounter certain problems with local health clinics, such as lack of attention, long waiting time, and dirty facilities.
DISCUSSIONS AND CONCLUSIONS
This multilevel analysis explores the effect of infrastructure provisioning on health service utilization and examines if governance shapes access with or without infrastructure provisioning. Results show that the presence of certain infrastructures, particularly health clinics and piped water, is crucial for enhancing people's health service utilization. However, people encounter different problems when receiving services at their local public clinics or hospitals. Results show that all governance indicators are significantly associated with the problems people encounter with their local clinics. Absent doctors, insufficient medicines and medical supplies, lack of attention, expensive services, long waiting time, dirty facilities, and such other problems that people often encounter with their local clinics are some examples of the lack of governance in the health sector in African countries.
There is some subnational variation in people's health service utilization, as service infrastructures are more concentrated in urban areas compared to rural areas. Results show that urban residents are more likely than rural residents to encounter certain problems with their local health clinics, such as lack of attention, long waiting time, and dirty facilities. These problems are more likely to be due to higher population density and greater service demand in urban areas than rural areas. However, these and other problems like lack of medical supplies, absent doctors, and expensive services are significantly related to public spending on health as well as lack of governance in the health sector. Improvements in health service utilization therefore require both better infrastructure provisioning—in terms of physical and capital stocks—and better governance, as the former does not guarantee the latter.
My results for country-level democratization provide partial support to Wullert and Williamson's (2016) findings that health services (in terms of child mortality) are worse in hybrid regimes than in both democracies and autocracies. However, Wullert and Williamson did not include an interaction term for democracy and GDP per capita, and I find that the model with the interaction term is more consistent with other models than the one with the democracy square term. My results suggest that democratization is most likely to enhance health service utilization in countries with higher levels of economic growth. This finding supports the classic claim made by Lipset (1959) and also supports Moon and Dixon (1985) as well as Mann's (1984, 2008) claims that infrastructurally powerful modern democratic states are better able to provide basic services to people. However, for democratization external pressure and conditional foreign aid may not work; as Wullert and Williamson (2016) warn, “Lacking internal willingness among elites to implement real change, democracy may flounder, resulting in political instability” (p. 10), which may lower people's health service utilization. The “real change” requires improving governance at all levels, as my analysis suggests. If state leaders are not committed to improving governance and enhancing people's service utilization, much public spending will be lost, as clearly evidenced in the infrastructure construction sector. Problems like absent doctors, insufficient medical supplies, lack of attention, expensive services, long waiting time, and bad working conditions at public clinics can be addressed without much spending if governance is improved in the health sector. Governments need to improve working conditions at the facility level and effectively monitor that the health infrastructure is functional and that the recruited staffs serve in their posts.
Appropriate state interventions are the developmental roles state leaders need to play to achieve developmental outcomes for a large cross section of people in society. For playing out developmental roles, state leaders need appropriate developmental structures—effective institutions and sufficient infrastructure provisioning. “Structures create the potential for action, playing out roles translates the potential into real effects,” as Evans (1995:77) claims. A combination of the developmental structures and developmental roles determines the levels of state capacity in achieving developmental success (Vu 2007). My analysis suggests that playing out developmental roles requires both a committed and effective government and adequate infrastructure provisioning.
My results for fiscal measures of state capacity in terms of state size point to the issue that Dawson (2010) raises. My analysis confirms his claim that nonfiscal measures of state capacity such as the rule of law (and additionally, control of corruption, democratic accountability, and government effectiveness) fare much better when our objective is to analyze the effect of state capacity on developmental outcomes, including better service provisioning and service utilization. States can be small for logistical reasons, as evidenced in Botswana and Rwanda, although Rwanda is an exceptional case where high state infrastructural power among the Hutu majority has been disastrous for the Tutsi minority (Mann 2005). In other cases, smaller states, measured in terms of population size, have a larger share of public consumption in GDP (Ades and Glaeser 1999; Alesina and Wacziarg 1998). This may happen because smaller countries have lower fixed costs in establishing a set of institutions to serve the population and because the costs of certain public goods, such as schools, clinics, roads, parks, libraries, and telecom infrastructures, “grow less than proportionally to the size of the population” (Alesina and Wacziarg 1998:308). However, this mechanism is completely different from the neoliberal claim that any increase in state size through tax revenue collection hampers economic growth and thereby service utilization.
Whereas the fiscal measures of state capacity focus on tax revenue collection and government spending as the main tasks of the state, the nonfiscal, governance-related measures of state capacity require a shift of focus to the expansion of human capabilities, including the enhancement of people's service utilization. By shifting the focus from development centered on economic growth and a free-market economy to human-centered development, states can better serve the people through developmental interventions and committed leadership (Ghobarah, Huth, and Russett 2004; Kaufman and Segura-Ubiergo 2001; Przeworski et al. 2000). However, developmental interventions are not only important at the country level but also much needed at the service utilization level, as this multilevel analysis shows. Whereas most ecological analyses rely on cross-national country-level data and thus miss the internal patterns and sources of variation in the results, this study accounts for those internal patterns and local sources of variation using hierarchical household survey data in a three-stage mixed-effects modeling. Failing to account for the within-country variation, cross-national ecological studies may offer erroneous theoretical insights into the underlying mechanisms through which country-level governance shapes individual-level developmental outcomes. To avoid this, my multilevel analysis illustrates how governance at different levels can affect outcomes and thus recommends that development scholars widen their focus beyond national-level governance and help policy makers identify at which level state interventions are most needed for removing barriers to development.
A final note of caution: although this study finds statistically significant correlations between infrastructure provisioning, governance, and individual-level health service utilization, the correlations may not suggest any causal relationship. For a causal analysis we need data over a considerable time period. Afrobarometer has the data for 16 countries over the period from 2002 to 2013, but PSU-level geographical information and a few other important variables are missing in the earlier surveys. For this reason, this study uses data from the most recent Afrobarometer surveys available for 34 African countries. Using the insights of this study, future studies may look into the causal mechanisms using appropriate longitudinal data as well as country-specific case studies.
- © 2017 by the Regents of the University of California. All rights reserved. Please direct all requests for permission to photocopy or reproduce article content through the University of California Press's Reprints and Permissions web page, www.ucpress.edu/journals.php?p=reprints.
Amm Quamruzzaman is a Course Lecturer of Sociology at McGill University. He is a development sociologist working on governance, gender quotas, democratic deepening, ethics in global health research, conflict and violence, inequality, and development. He has written The Militia Movement in Bangladesh: Ideology, Motivation, Mobilization, Organization, and Ritual (Saarbrücken, 2010), Project Chittagong Hill Tracts: Field Notes (coauthored with UK Committee for UNICEF/UK, London, 2006), and articles on education policy, female political representation, governance for development, happiness, and violence.