Blog post by Daniel Armah-Attoh
Ghana’s place at the forefront of African democracy and good governance has been called into question by a recent series of corruption scandals. Quite dishearteningly, some public officials have been found defending alleged wrongdoers in media discussion programs, and some whistle-blowers suffered reprisals instead of being protected.
Within this context of loss of faith in government's political will to fight the canker of corruption, the Center for Democratic Development (CDD-Ghana) in December 2014 released the 2014 Afrobarometer Round 6 findings on Ghanaians' perception of corruption. A day later, the Ghana Integrity Initiative (GII), the local chapter of Transparency International (TI), released the TI Corruption Perceptions Index (CPI) 2014 for Ghana.
Hard facts vs. unpalatable results
Instead of examining and using the findings of Afrobarometer and TI, many government and ruling-party officials, along with media persons close to them, chose to attack the credibility of Afrobarometer’s survey.
They argued that the sample size of 2,400 was too small for national-level generalization and that the distribution of the sample was skewed to favour the opposition political party.
Certainly, no research project is above criticism. However, such criticism should be underpinned by hard facts and an understanding of survey methodologies, rather than an unwillingness to hear unpalatable results.
Perceptions of corruption
Without taking you through a university course in statistics, let’s highlight a few facts about how Afrobarometer and TI used different methodologies to arrive at very similar conclusions about the high levels of perceived corruption in Ghana.
Corruption is illegal and therefore generally hidden and difficult to gauge. Perceptions of corruption are one way to estimate the real situation. Both Afrobarometer and TI findings are perception-based. But the two gather perceptions from different categories of people: Afrobarometer asks ordinary citizens, while TI asks experts.
Random selection and proportionality
Afrobarometer surveys adults selected at random. This does not mean they are chosen haphazardly, by standing on a streetcorner and talking to passers-by who are willing to talk. It means they are chosen in accordance with a scientific method that eliminates the possibility of bias by giving every adult Ghanaian the same probability of being interviewed. (You can read the details of Afrobarometer’s selection process in our survey manual)
The selection of respondents is randomized at every step, from the random selection of 300 enumeration areas across the regions of Ghana to the random selection of eight households in each enumeration area and the random selection of one person in each selected household. This method also ensures that every region is represented in exact proportion to its share of the country’s adult population. The same is true of urban vs. rural areas. And exactly 50% of respondents are men, and 50% are women.
Because of this proportionality, it is impossible for the sample to be skewed in favour of certain regions. If Greater Accra contains 19% of Ghana’s adult population, with 92% of those being urbanites and 8% rural residents, then 19% of the Afrobarometer sample will be from Greater Accra, with the same urban-rural breakdown. Critics who argued that the Afrobarometer sample was skewed to favour the main opposition party because 31% of survey respondents were selected from the two regions where the main opposition party won the 2012 elections (i.e. 20% from Ashanti and 11% from Eastern) neglected to consider that those two regions represent exactly the same proportion of the country’s adult population. By the same token, 69% of the Afrobarometer sample came from the other eight regions, where the current government won the 2012 elections – because 69% of the country’s adult population lives in those regions.
Sampling is like a spoonful of soup
The critics’ other argument – that a sample of 2,400 respondents is too small to represent all of Ghana – is equally flawed but requires a longer response.
Some critics seem to mistrust the idea of sampling in general. Maybe thinking of all the sampling we do in our everyday lives will alleviate their suspicions. If we’re preparing a big pot of soup, and we want to know whether there’s enough salt, or too much ginger, we don’t drink the whole pot of soup. We taste a spoonful – a sample. Or maybe two, just to be sure. Or three, if it’s really tasty.
Similarly, if the doctor wants to test our blood to find out what ails us, we don’t ask him or her to draw all of our blood, just to be sure s/he gets the diagnosis right.
More sophisticated critics of sampling will need a more sophisticated answer.
To know exactly how all Ghanaians see things, you have to talk to every Ghanaian. A census is a great approach – if you have unlimited money and nothing better to do.
Increasing sample size makes small gain in precision
If money, human resources, and time are not unlimited – i.e. in the real world – you calculate a sample size for your survey that will be precise enough for your purposes and still affordable. A survey’s level of precision is expressed as a margin of error. Afrobarometer surveys with samples of 2,400, as in Ghana, have a margin of error of +/-2% with a 95% confidence level. This means that if you conducted the same survey under the same circumstances 100 times, 95 times you would have the same finding, plus or minus 2%.
We could improve this level of precision by increasing the sample size. For example, if we interviewed 4,800 people instead of 2,400, we could reduce the margin of error from +/-2% to +/-1.5%. But we would also be doubling our costs – and for a fairly small gain in precision. For our purposes, it suffices to be able to report that 89% +/-2% (i.e. between 87% and 91%, about nine out of 10, an overwhelming majority) of Ghanaians say that "some", "most", or "all" police are corrupt, rather than spending twice as much money just so we can report that between 87.5% and 90.5% of citizens say so.
In sum, Afrobarometer surveys have the strengths and limitations of the best survey methodologies – reviewed, endorsed, and applied by the scientific community throughout Africa and the world.
Corruption Perceptions Index
The Transparency International (TI) Corruption Perceptions Index (CPI), on the other hand, is a composite constructed from corruption-related data produced by other independent bodies, informed by evaluations by business people and experts/analysts of the degree of corruption within the public sector in each country. A country's CPI score is calculated as an average of all the available scores for that country. TI’s 2014 CPI score for Ghana was based on data from eight reputable sources:
- Bertelsmann Foundation Transformation Index (2014)
- Economist Intelligence Unit Country Risk Ratings (2014)
- Global Insight Country Risk Ratings (2014)
- Political Risk Services International Country Risk Guide (2014)
- World Economic Forum Executive Opinion Survey (2014)
- World Justice Project Rule of Law Index (2014)
- World Bank Country Performance and Institutional Assessment (2013)
- African Development Bank Governance Ratings (2013).
It may interest critics to know that some of these sources used by TI also depend on Afrobarometer for information. The World Bank Country Performance and Institutional Assessment uses information from the Worldwide Governance Indicators, which includes Afrobarometer among its nine data sources. The African Development Bank governance ratings also use Afrobarometer data as one of its sources of evidence. Though Bertelsmann does not use opinion data in the computation of its index, it often cites Afrobarometer findings in its country reports.
Different methods, similar conclusion
Given the very different methodologies used by Afrobarometer and TI, what is remarkable is the degree of convergence between their findings in 2014.
Afrobarometer found that large majorities of Ghanaians (ranging from 83% to 89%) perceive “some,” “most,” or “all” police officers, national government officials, parliamentarians, judges and magistrates, tax officials of the Ghana Revenue Authority, chief executives of local government bodies, president and officials in his office, local government representatives, and officials of the country's electoral management body to be involved in corruption (Figure 1). The 2014 figures are marginally lower than those from the 2012 survey for all the listed public officials.
Figure 1: Afrobarometer corruption perceptions | 2012-2014
Respondents were asked: How many of the following people do you think are involved in corruption, or haven’t you heard enough about them to say: (a) the president and officials in his office? (b) members of Parliament? (c) Electoral Commission officials? (d) government officials? (e) assembly men and women? (f) district chief executives? (g) police? (h) tax officials (i.e. Ghana Revenue Authority officials)? (h) judges and magistrates? (% saying “some,” “most,” or “all” are corrupt).
Note: Corruption among officials of the Electoral Commission officials was not asked in 2012.
Afrobarometer’s findings that things are bad, but at least slightly improved, are consistent with TI’s findings. TI gave Ghana a CPI score of 48, based on a scale of 0 (highly corrupt) to 100 (clean). This score placed Ghana at No. 61 out of 174 countries assessed (with No. 1 being the least corrupt) (see https://www.transparency.org/cpi2014/results). Comparing Ghana's 2014 score to those of 2013 (46) and 2012 (45), the country improved marginally.
Both Afrobarometer and TI thus reported marginal improvements, but with the overall conclusion that corruption is still a major challenge facing the country. Both cited high-profile corruption scandals among contextual issues.
Both organisations are supported by prestigious international donors, adding further to their credibility. Those attacking or ignoring Afrobarometer and TI because their findings are politically unpalatable are doing their country more harm than good. A better approach would be to critically consider research findings when formulating public policy.