- Top of page
Corruption is a sensitive subject that may be an issue in all nations to some extent albeit rich or poor (United Nations Office on Drugs & Crime 2009). It is a worldwide problem that can widen the socio-economic gap within society (Gupta et al. 2002), may have a negative influence on socio-economic development (Assiotis & Sylwester 2010; Mauro 1995; Méndez & Sepúlveda 2012) and can steer government spending (Shleifer & Vishny 1993; Mauro 1998).
There is a growing body of literature linking higher societal corruption to negative health outcomes and a reduction in the use of public health services (Azfar & Gurgur 2008). According to the social determinants of health models, both the person and the environment in which a person lives can influence their health status (Arah 2009). Thus, if a person is living in a highly corrupt society, then this might have an impact on his or her health status. It is likely that vulnerable groups might feel the impact of corruption most (Savedoff & Hussmann 2006). For example, the practice that doctors charge clients for services or prescription medicine that should have been rendered free (Ferrinho et al. 2004) could disproportionately be an issue for poorer populations who are unable to pay for treatment or other health services. Containing corruption has been a recognised long-standing problem in many middle- and low-income parts of the world. Africa in particular has had the goal of eradicating corruption on the agenda for quite some time, and there are many initiatives to combat corruption (United Nations Economic Commissions for Africa 2009). Some progress has been made, but change is sluggish. Many African governments make plans to rectify their regimes, but these are often marred by corrupt elements, and as a result, anticorruption policies fail to be enforced (Kpundeh & Dininio 2006). Studying the population health consequences of corruption on the health of the population may provide another argument for a corruption-free society. Furthermore, African countries differ in the extent of corruption (United Nations Economic Commissions for Africa 2009); thus, we wished to determine whether such differences are associated with varying levels of population health.
Usually studies on corruption are based on single-country studies. Only a handful of public health studies empirically investigate the health consequences of corruption in an internationally comparative manner, and those few studies often utilise aggregated child population health data (Hanf et al. 2011). Findings consistently point towards corruption being associated with child health outcomes (Transparency International 2006; Hanf et al. 2011). So far, there is very little empirical evidence that has examined the association of societal corruption with adult general health using a comparative approach (Lewis 2006).
In this article, we study 20 continental and island African countries. Using the social determinants model as a guide, we view national corruption as a country-level characteristic that may have an influence on health, and might be, thus, a possible contextual-level determinant of health. Given this perspective, we examine the extent to which perceived national corruption influences general health while accounting for individual- and country-level socio-economic development and religion. We also investigate the relationship between corruption and general health according to sex, age and social status (measured by education and occupation). We expect to find the impact of corruption, if any, to be largest among lower socioeconomic groups.
- Top of page
Our data set was derived from the latest WHO cross-national comparative World Health Survey, which began data collection during the years 2000–2004, and has data from 70 countries (World Health Organization 2010). Detailed information regarding the survey creation process, data compilation, response analyses and country information is available via the WHO website (World Health Organization 2010). Two surveys were implemented, one pertaining to the household and the other containing individual health questions. We extracted individual cross-sectional health data from 20 African nations surveyed by the WHO. Respondents were excluded if missing information on the variables utilised in this study. This equated to 72 524 men and women being included in this investigation.
Our outcome variable of interest was self-reported (general) poor health. We created this dependent variable by extracting responses from the question ‘In general, how would you rate your health today?’ Five answer choices were available to choose from, namely ‘very good’, ‘good’, ‘moderate’, ‘bad’ or ‘very bad.’ We transformed these choices into binary variables by grouping the responses ‘very good’ and ‘good’ as 0 and ‘moderate’ ‘bad’ and ‘very bad’ as 1, with the former being classified as the reference group.
Our main country-level variable of interest was corruption. Corruption has many facets, making it a tremendously difficult term to define, identify and measure (United Nations Economic Commissions for Africa 2009). The World Bank describes corruption as manifesting in both ‘overt’ and ‘covert’ forms (The World Bank 2010). Usually, overt (i.e. the blatant form of) corruption makes headline news. It refers to the ‘abuse of public office for private gain’ (The World Bank 2010). Covert corruption is often undetectable, but is so interwoven into the seams of society that it goes unnoticed or occurs so often that it is considered commonplace. The World Bank describes this type of corruption as being carried out by those in service positions (doctors, teachers, inspectors and so on; The World Bank 2010).
In this article, we used the corruption perception index (CPI) of Transparency International (Transparency International 2006). Transparency International is an organisation that monitors and works to eradicate corruption within societies worldwide. Transparency International defines corruption as ‘the abuse of public office for private gain’ (Transparency International 2006). The purpose of the CPI is to measure ‘both administrative and political corruption’ (Transparency International 2006). To construct the index, people who live within and outside of the country completed questionnaires. These questionnaires accessed their perception of national corruption levels (Transparency International 2006). Questions regarding ‘bribery of public officials, kickbacks in public procurement, embezzlement of public funds etc.,’ were asked (Transparency International 2006). The CPI was first created in 1995 and included 41 countries worldwide: today it incorporates 183. We extracted figures from the year 2007. This year was selected because it was the earliest available time period that included all 20 African countries used in this analysis. The CPI is measured on a 0–10 scale, with a higher score pertaining to less perceived corruption within society (Transparency International 2006). In this study, we inversed the CPI score, so that a higher score amounted to more perceived national corruption.
Individual-level variables included sex, age (as a continuous variable) and educational attainment (i.e. no formal schooling to postgraduate education completed) and occupational status (i.e. eight categories ranging from non-manual and manual occupations to the unemployed).
To account for the economic situation of countries, we included two contextual-level human development factors, namely gross national income (GNI) and the mean years of schooling for adults aged 25 and older. These data were extracted from the United Nations Human Development Reports for the year 2009 (United Nations 2012). For statistical purposes, the GNI variable was divided by 1000 before inclusion into the analysis.
Africa is a highly diverse, religious continent with a large Muslim population in northern countries. We control for religion to account for the possibility that the prevalence of poor health may differ between predominantly Muslim countries and other countries that have a smaller Muslim population. To create the religion variable, we classified each country using a binary variable that assigned 1 to countries identified by the Pew Research Center (Pew Research Center 2009) as 60% or more Muslim and a 0 to those countries with less than 60% Muslim population. As a sensitivity analysis, we also examined the influence of using the religion variable as continuous Muslim country percentages in statistical analysis.
To investigate the extent to which national corruption could be associated with general poor health, we conducted multivariable analyses using several random-intercept multilevel logistic models. Multilevel analysis is appropriate for our hierarchically structured data that were created by merging individual-level data with country-level data. It also appropriately accounts for the clustering of outcomes within countries and for between-country differences (Greenland 2000). We constructed three multilevel models. The first model regressed the outcome on CPI while adjusting for age and sex only. Following this, model 2 further adjusted for individual-level educational attainment and occupational status. Lastly, in model 3, country-level information on mean years of schooling, gross national income and religion were included.
Furthermore, we conducted stratified analyses for gender, age, educational level and social status adjusting for individual- and contextual-level variables in the same manner as the previously discussed models 1–3. Given the nature of our study's objective, only the results for corruption are shown. For the age-stratified analysis, the cut-off point was the mean age (i.e. 38 years). The education stratification involved constructing a ‘high’ vs. ‘low’ category for educational attainment. Those who completed high school or more were placed in the highly educated group, while all others were classified in the low education category. We separated social status into ‘non-manual’ vs. ‘rest’ using the occupational status variable. All white-collar functions were placed in the non-manual group. We conducted all analyses in Stata 11.1 (College Station, TX).
- Top of page
Table 1 shows the descriptive results. For both women and men, the highest prevalence of poor health was found in Swaziland, 72.4–70.9% respectively. Perceived corruption rates ranged from 1.9 (South Africa) to 5.2 (Chad). Of all the countries, eight of 20 identified as having a 60% or more Muslim population. The mean GNI was USD 2859.69 (SD = USD 3139.56). The mean years of schooling was 4.8 years (SD = 2.3 years). More than half (78.1%) of the studied population had no formal education or only primary school education.
Table 1. Descriptive information of study participants from the World Health Survey a–d
|Country|| n ||% Poor health (women)||% Poor health (men)||Corruption perception||% Muslim|
Table 2 contains the results of the multilevel models. In the first model (model 1), we adjusted for age and gender. We observed that higher levels of national corruption appeared to indicate a higher prevalence of poor health in the total population, although the confidence intervals include 1 (OR = 1.12, 95% CI: 0.87–1.43). Controlling for individual-level socio-economic factors (model 2) showed a similar finding as in model 1. Adjusting for all contextual-level variables in model 3 increased odds ratio for the association between corruption and poor health to 1.62 (95% CI: 1.01–2.60). We also adjusted for religion as a continuous variable. Overall results were similar to using the categorical religion variable; therefore, we opted to not show continuous results.
Table 2. Odds ratios and 95% confidence intervals for the associations between perceived corruption and self-reported poor health, adjusted for individual- and country-level factors a,b
| ||Model 1 + age and sex||Model 2 + education and occupation||Model 3 + contextual-level indicators|
|OR||95% CI||OR||95% CI||OR||95% CI|
|GNI|| || || || ||1.10||0.97–1.24|
|Mean years of schooling|| || || || ||1.09||0.95–1.25|
|<60% Muslim|| || || || ||1||Reference|
|>60% Muslim|| || || || ||1.43||0.85–2.41|
|Education|| || || || || || |
|Postgraduate degree|| || ||1||Reference||1||Reference|
|College graduate|| || ||1.23||0.99–1.52||1.23||0.99–1.52|
|High school graduate|| || ||1.36||1.11–1.67||1.36||1.11–1.67|
|Secondary school completed|| || ||1.38||1.13–1.69||1.38||1.13–1.69|
|Primary school completed|| || ||1.57||1.29–1.91||1.57||1.29–1.91|
|Less than primary school|| || ||1.75||1.43–2.14||1.75||1.43–2.14|
|No formal schooling|| || ||1.63||1.33–1.98||1.63||1.33–1.98|
|White collar|| || ||1||Reference||1||Reference|
|Armed forces|| || ||1.04||0.85–1.27||1.04||0.85–1.27|
|Clerk|| || ||1.03||0.88–1.20||1.03||0.88–1.20|
|Service or sales worker|| || ||1.10||0.99–1.23||1.10||0.99–1.23|
|Agricultural or fishery worker|| || ||1.23||1.12–1.35||1.23||1.12–1.35|
|Craft or trades worker|| || ||1.09||0.98–1.22||1.09||0.98–1.22|
|Plant/machine operator|| || ||1.21||1.07–1.37||1.21||1.07–1.37|
|Elementary worker|| || ||1.21||1.09–1.35||1.21||1.09–1.35|
|Not working for pay|| || ||1.49||1.37–1.62||1.49||1.37–1.62|
Table 3 shows the multilevel stratified analysis for gender, age, educational status and social status in addition to the significance test for the effect modification. Similar relationships were observed for all subgroups, with higher national corruption appearing to be associated with an increase in poor health prevalence. While figures appear virtually identical for all subgroups, as compared to their counterparts, the odds ratios were slightly higher in the final model (model 3) for men (OR = 1.66, 95% CI: 1.03–2.67) than women (1.60, 95% CI: 0.99–2.57) and for younger adults (OR = 1.67, 95% CI: 0.99–2.82) than older adults (OR = 1.55, 95% CI: 1.02–2.34). The odds ratios were also higher for less educated (OR = 1.61, 95% CI: 1.01–2.58) and the rest occupation category (OR = 1.62, 95% CI: 1.01–2.59) than for better educated and non-manual occupation respondents (OR = 1.40, 95% CI: 0.83–2.37; OR = 1.56 95% CI 0.93–2.62, respectively). For all these categories, the effect modifications by the stratified variables were significant.
Table 3. Stratified analyses (i–iv) showing odds ratios and 95% confidence intervals for the associations between perceived corruption and self-reported poor health, adjusted for individual and country-level factorsa–d
| ||Model 1 + age and sex||Model 2 + education and occupation||Model 3 + contextual-level|
|OR||95% CI||OR||95% CI||OR||95% CI|
|(i) Gender stratification|
|P value||<0.001|| ||<0.001|| ||<0.001|| |
|(ii) Age stratification|
|Age, <38 Only||1.16||0.88–1.51||1.13||0.86–1.49||1.67||0.99–2.82|
|Age, >38 Only||1.08||0.86–1.36||1.07||0.85–1.34||1.55||1.02–2.34|
|P value||<0.001|| ||<0.001|| ||<0.001|| |
|(iii) Educational stratification|
|P value||<0.001|| ||<0.001|| ||<0.001|| |
|(iv) Occupational stratification|
|P value||<0.001|| ||<0.001|| ||<0.001|| |
- Top of page
This study is one of the first investigations to explore perceived national corruption in a multitude of African countries merging contextual- and individual-level health data on several thousands of adults. We observed that a higher level of perceived national corruption in society is associated with poorer population health in both women and men. National perceived corruption levels seem to be associated with negative health outcomes in both young and old, where the lower social status groups seem to fare slightly worse. This association became more pronounced after adjustments for national-level socio-economic factors and religion.
Our study has limitations. First, we caution that this study uses data of 20 African countries. As a result, the extent to which the results can be generalised to other parts of the world is uncertain. Future research should be geared towards investigating this subject while incorporating other countries around the world. A second concern is the measure of self-reported poor health: The self-reports are potentially sensitive to personal and cultural characteristics (Sen 2002). We recognise the shortcomings of using self-reported health as an outcome. However, it is a frequently used indicator in the health sciences, and studies suggest that self-reports of general health correlate highly with subsequent mortality (Idler & Kasl 1991). Given that this is a comparative study, there is a particular concern with the comparability of the self-reported health indicator across countries. Evidence suggests that there might be differences between countries and social statuses regarding how items on self-reports of health are answered (Sen 2002). The WHO designed the World Health Survey using extensive methodological techniques to account for cross-country differences (Üstün et al. 2010). Even though measures have been taken, our results should be interpreted with caution.
Third, there is the issue of the corruption measure CPI, which is widely used in the field of economics, and which has been employed to measure national corruption worldwide (Andersson & Heywood 2009). However, the CPI has natural limitations, and there is controversy surrounding its use (Mauro 1995). First, the CPI is based on subjective response data (Johnston 2000), which renders it sensitive to response shift issues. Responses might change depending upon pre-established conceptions respondents may or may not have about the country before taking part in the survey (Mauro 1995). Second, the CPI is an aggregate indicator of corruption and does not specifically measure the many different forms of corruption (Mauro 1995, 1998). Third, the CPI measures structural flaws in the system and only classifies the ‘administrative and political corruption’ (Transparency International 2006). Perhaps other forms of corruption might have a different impact on health. Despite these possibilities, the CPI corresponds to other corruption indexes utilising different surveys and methodological techniques (Wei 1998). Our analysis is nonetheless simply a first step; future research should expand on these findings by accounting for other forms of corruption.
Although we observed that corruption might be associated with negative health outcomes, given the core (mostly untestable) conditions required for identifying and estimating causal effects from observational studies, our study does not allow us to draw any causal conclusions (Arah 2008; Pearl 2009; Arah et al. 2013). Furthermore, it is relevant to consider the possibility of reverse causality or the effect of population health on the prevalence or perception on corruption. To measure the possible influence this might have had on our study, we would need to have reliable and valid longitudinal measurements of both corruption and health outcomes with appropriate lag periods between corruption and health measures (Berkman & Kawachi 2002). Without this longitudinal data, our ability to investigate the impact of possible reverse causality is limited.
Social determinants of health are factors related to a person's economic and social situation, which influence their health (Gupta et al. 2002). As previous research has shown, health status can also be influenced by the contextual or environmental situation of each country (Shleifer & Vishny 1993; Mauro 1995, 1998; Méndez & Sepúlveda 2012). In our study, we observed that a higher perceived national corruption seemed to be associated with a much higher prevalence of poor health after accounting for contextual-level socio-economic factors and religion. This suggests that the higher perceived national corruption in African societies, the poorer the health of the total population. To a certain extent, this was irrespective of age, sex and social status. These results are consistent with other contextual or environmental studies that found higher levels of corruption to be linked with poorer health outcomes; these include a higher child and maternal mortality and lower life expectancy (Azfar & Gurgur 2008; Hanf et al. 2011; Holmberg & Rothstein 2011). In addition, our findings buttress those of an earlier work that introduced the inequity-in-health index, showing that the index increased (hence, became worse) with increasing corruption (Eslava-Schmalbach et al. 2008).
Several mechanisms are evident that may explain the relationship between corruption and health. In economics, for example, plenty of evidence has been brought to light that corruption negatively influences development, including education (Mauro 1995; Azfar & Gurgur 2008). Low education levels in society, especially among women, will have a negative impact on health (United Nations 2003). However, we found an association with corruption also after controlling for education level. Corruption in the healthcare industry is also a major concern in many poor countries: If health professionals are corrupt, they cause access problems, and this will have negative health consequences (Holmberg & Rothstein 2011). Corruption has been named as the number one reason behind children dying of malaria in Tanzania, as corrupt practices prevented them from receiving proper medical care (The World Bank 2010).
Prior reports claim that corruption has an unbalanced negative impact on women compared with men (United Nations Development Program 2012). For instance, women come more frequently into contact with social services than men, and if this sector of government is corrupt, then women will feel the impact of this inadequacy more than men (Hossain & Nyamu-Musembi 2010). Women in low- and middle-income countries are over-represented in lower-paid positions, and therefore, they have limited financial means to pay bribes (Hossain & Nyamu-Musembi 2010). As a result, they may more often be unable to access health services (Hossain & Nyamu-Musembi 2010). Although the aforementioned examples highlight the different pathways in which the two sexes may be unequally confronted by corruption within society, our results suggest that, from a health perspective, a stronger association for women than for men is not evident.
Our findings suggest that lower social classes might carry a heavier burden when living in a society rife with corrupt elements. The literature suggests that corruption has a marked impact on poor populations, especially within Africa (Gyimah-Brempong 2002; Transparency International 2006). This might be related to economics, as poor people often lack the financial resources to cope with corruption (Teggemann 2013). For example, bribing in the healthcare sector is a serious problem across much of Africa and can have serious health consequences for the poor.
Corruption in the African context is a complex matter. It can be part of a vicious circle that is difficult to amend. If the aim is to promote population health and move a step forward in achieving MDGs, it might be necessary to restructure governmental regimes marred by corrupt elements. These results provide ammunition for those fighting for a corruption-free African society. When national corruption is a problem, quite possibly, everyone's health suffers.