Empirical evidence of the gender dimensions of bribery in Ghana

3.1. Definition of concepts

3.2. Data source

The data presented in this report was collected in the 2021 Ghana Integrity of Public Services Survey (GIPSS), a project funded by the German Federal Ministry of Cooperation and implemented by United Nations Office on Drugs and Crime (UNODC) and the Commission on Human Rights and Administrative Justice (CHRAJ) in collaboration with Ghana Statistical Service (GSS). This survey is the first comprehensive, nationally representative population survey on corruption in Ghana, including 15,000 respondents across the country.

Accurate data collection is essential for unbiased results, especially when considering a sensitive topic such as corruption where there is a particular risk due to non-responses. In this study, there are also additional concerns with differences in the experience or perception of bribery between genders, which could bias the results. Bias could also arise if there were differences in the response rate, or the likelihood of responding to certain questions, between men and women.

In this study data was collected from 1,500 Enumeration Areas (EAs), which were selected with probability proportional to size. In each EA, a fixed number of 100 households were selected, from which one adult member was selected for the interview. No replacements and no changes in the preselected households were allowed in order to prevent bias. In total 14,843 (99.0%) interviews were conducted, resulting in a nearly perfect response rate for both men and women. In addition, every respondent had to answer all questions. Hence, in the overall response rate, as well as question-specific response rates, there are no significant differences between men and women, making it less likely that differences in response rates affect the results.

The survey covered 23 different types of public officials, ranging from health workers and teachers to police officers and judges, and asked respondents about their contact, payment, and refusal of bribes, as well as their knowledge of anti-corruption agencies and their exposure to other forms of corruption, such as nepotism and vote buying. The survey also included questions on sexual favours.

3.3. Estimation technique

A logistic regression model is used to compute the likelihood of a person paying a bribe, conditional on the set of covariates discussed above. The same model is used to predict the likelihood to refuse to pay a bribe, as well as to predict the likelihood of having received a request for a sexual favour.

Given that the survey focuses on individual-level experiences of corruption, with only one respondent per household, the data is clustered at the individual level within EAs. Responses within the same EA may be correlated due to shared contextual factors, such as local governance or socioeconomic conditions. To account for this, clustering was specified at two levels: EAs as the primary sampling units and individuals within EAs. This ensures robust standard errors that reflect the hierarchical structure of the data and potential intra-EA correlation in corruption experiences.

To address any concerns with the size of the survey weight affecting the results, the same models were run with and without applying the survey weight as a robustness check. The results showed no significant differences in the overall outcomes, with no changes in significance or substantial size differences in the main coefficients. The only exception was the significance of contact with only a female official, which was significant at a 10 percent level when weights were included, but not when excluded. This is likely due to the small number of respondents who encountered only a female official, making the result more sensitive to the weights. This however does not alter the main finding that contact with male officials significantly increases the likelihood of paying a bribe. Overall, the inclusion of survey weights does not affect the main conclusions of the paper, confirming the robustness of the results.

OPEN IN VIEWER

Table 2.

The impact of gender in the logistic regression on paying and refusing to pay bribes.

	Odds Ratio
	Paying	Refusing to
Covariate	a bribe	pay a bribe
Base category (Female)
Male	1.27***	1.04
	(0.00)	1.04 (0.68)

Including all other correlates, not displayed in this table.

Clustered by enumeration area and individual, p-values in parentheses.

Signif. Codes: ***: 0.01, **: 0.05, *: 0.1

OPEN IN VIEWER

Table 3.

The impact of gender of public officials in the logistic regression on paying and refusing to pay bribes.

	Odds Ratio
	Paying	Refusing to
Covariate	a bribe	pay a bribe
Base category
(contact with male and female officials)
Contact with a male official	1.92***	0.98
	(0.00)	(0.87)
Contact with a female official	1.19*	0.96
	(0.10)	(0.84)

Including all other correlates, not displayed in this table.

Clustered by enumeration area and individual, p-values in parentheses.

Signif. Codes: ***: 0.01, **: 0.05, *: 0.1

For the regression on paying bribes, the subset of those who have been in contact with the public officials was used. However, for the regression on refusing to pay a bribe, only the sub sample of those who either paid or refused to pay a bribe are included.

4.1. Contact with public officials by gender

In line with GSS and UNODC, ¹ being involved in bribery can be defined as both paying a bribe as well as refusing to pay a bribe. Using that definition, 28.9 percent of men was involved in bribery, and 17.2 percent of women was involved in bribery. However, that most simple definition is not a great indicator, as it does not account for the fact whether people have even been in contact with public officials. Hence, one needs to take a step back and first analyse how many people are in contact with public officials. Then, as only those who have been in contact with public officials have had the chance to be involved in bribery, the prevalence of bribery can be computed for those who have been in contact with public officials.

In Ghana in 2021, 83.8 percent of the adult population had been in contact with at least one public official. There was a slight difference between men and women, where 84.9 percent of all men had been in contact with at least one public official, with a respective 82.9 percent of all women. In addition to contact in general, it is informative to see how the contact rate with specific types of public officials is correlated with gender. ¹ highlights that that one of the most important factors when analysing the differences in bribery experiences between the two sexes is the type of officials there are in contact with. The contact rate with different public official types by gender display some remarkable findings, as presented in Table 1.

Table 1 shows that for most types of officials the contact rate is higher for men, particularly for police officers, local government representatives and officials from the Driver and Vehicle Licensing Authority. On the other hand, women have higher contact rates with doctors, nurses and midwives, as well as with other health workers in public hospitals.

This result directly indicates the necessity of this paper. Since there is a clear difference between the type of officials men and women interact with, it is insufficient to ignore this fact and continue to give descriptive statistics about which gender is more prevalent to pay bribes, without controlling for the type of official they interact with. As it could be that it is the official type people are in contact with, rather then the gender of people, driving differences in bribery prevalence between men and women.

4.2. Prevalence of bribery by gender

Even in a case where differences in prevalence of bribery between men and women is driven by other causes, it remains insightful to see how that results in differences in bribery prevalence between genders. The headline figure presented in the gender dimensions chapter of the Ghana corruption report ¹ is that men are 1.6 times more likely than women to pay or be asked to pay a bribe when interacting with public officials: 34.1 percent of men pays or is asked to pay a bribe when interacting with public officials, as opposed to 20.7 percent of women. This is a very large difference, which makes it worthwhile diving deeper in understanding the dynamics of prevalence of bribery by gender.

Men and women exhibit notable differences in bribery behaviour, with 29.9 percent of men and 18.1 percent of women having paid a bribe when in contact with a public official. However, when considering refusing to pay bribes the differences were not as stark, 12.2 percent of men and 12.7 percent of women refused to pay a bribe when having been solicited to pay a bribe. This study emphasises the importance of distinguishing between payment and refusal of bribes, as detailed questions on official type and gender of the official are separately answered for paying and refusing bribes.

One explanation of the differences in the bribery prevalence could be that men and women have different perceptions about bribery. However, the survey outcomes show that there are no major differences in the perceptions of men and women about bribery. ¹ shows that around 80 percent (80.3 for men, 80.5 for women) of people says it is never acceptable for a public official to request a bribe in order to speed up administrative procedures. Moreover, men and women also gave similar answers to questions about the perceived frequency of bribery requests by public officials. Hence, the perceptions about bribery cannot be a direct explanation to the major differences that are found.

4.3. Bribery by gender whilst controlling for other variables

To be able to give more insight into the drivers of bribery, it is essential to return to other factors that may influence the prevalence of bribery. A great advantage of the Ghana Integrity of Public Services Survey 2021 is that most of the questions concerning bribery were asked separately for 23 different official types, and many social and demographic variables were also collected. This allows for testing the difference between men and women and their tendency to pay bribes, controlling for a wide range of other variables which may influence the likelihood of a bribe occurring.

In Section 3.4, the selection of the correlates used in the logistic regression is justified. In summary, a host of controls that correlate with the likelihood to pay bribes are included, in doing so it is possible to test the role of gender keeping all other variables constant.

When controlling for the wide range of socio-demographic variables, knowledge of anti-corruption agencies, the frequency of contact, the type of official, and the sex of the official, it is possible to see whether or not there is still a significant gender dimension to the prevalence of bribery. Table 2 shows the odds ratio of the logistic regression model of the gender variable, whilst controlling for all other variables. The full regression table can be found in Table A1 in the Appendix.

From Table 2 it can be concluded that keeping all other factors constant, men are 27 percent more likely to pay a bribe when they are in contact with a public official as opposed to women, which is statistically significant at the 1 percent level. On the other hand, there is no statistical difference between men and women in their likelihood to refuse to pay a bribe when in contact with a public official, when controlling for all other variables.

This means that there is no significant gender dimension to the refusal to pay bribes. This requires more research into specific reasons and mechanisms behind refusing to pay bribes. However, the main insight from this analysis is that when controlling for all the other variables, men are still 27 percent more likely to pay bribes than women.

From this section three major conclusions can be drawn. First of all, men are slightly more often in contact with public officials than women, and especially the type of officials that men and women are in contact with differs by gender. Secondly, it can be seen that men are more likely to pay bribes than women, while the perceptions of bribery is not different between the genders. Third, when controlling for other variables, there still remains a statistically significant difference between the likelihood of men and women to pay bribes.

When comparing the headline figure of GSS and UNODC ¹ that men are 1.6 times more likely to bribe than women, to our more thorough econometric analysis, the result is that (even when taking the official type, socio-demographic characteristics and some additional variables into account) men are still significantly more likely to pay bribes than women, but men are only 1.27 times (27%) more likely to pay bribes then women. This result shows that there are some variables, such as the type of official people are in contact with, that explain part of the disparity between the bribery prevalence of men and women, but not all differences can be accounted for. This finding is essential as proof that there are still unexplained differences between men and women in the context of bribery in Ghana. Hence, leading to the direct conclusions that additional research is still required on what drives these gender disparities in bribery.

5.1. Bribery by gender of public officials

The headline figure of GSS and UNODC ¹ on the differences between male and female public officials is: “The share of male public officials who solicit or take a bribe when interacting with the public is roughly twice that of female public officials”. It is presented that 25.2 percent of male public officials solicits or takes bribes as compared to 13.1 percent of the female public officials.

Now let us shortly digress into the public perception of the role of the gender of the public official. Public perceptions on the role of the gender of public officials indicate that 60.4 percent of Ghanaians believe male officials are more likely to request bribes. Men, at 62.8 percent lean slightly more towards this view compared to 58.5 percent of women. The perception is rooted in the belief that men, driven by a greater responsibility to provide for their families, are less hesitant to solicit bribes. It’s essential to note that these are public perceptions, not reflective of the actual behaviour of officials, highlighting the need for further research in this area.

Getting back to analysing bribery by the gender of public officials, the same problems arise as when analysing the role of the gender of the bribe payer: what if it is not really the gender of the official that makes the difference, but some other variable which is correlated with the gender of officials? One step towards analysing this difference was made by GSS and UNODC, ¹ by analysing the prevalence of bribery by the sex of the public official and the type of public official. This shows that under police officers, 52.8 percent of the male public officials request bribes, as compared to a much lower 27.3 percent of female police officers. This insight shows that even when controlling for the type of official, significant gender disparities remain.

5.2. Bribery by gender of public official whilst controlling for other variables

Similarly as in Section 4.3, in this Section the logistic regressions on paying and refusing to pay bribes is analysed, but now the focus is shifted to the estimates on the effect of the gender of the public official. These results are presented in Table 3.

Table 3 shows that when controlling for other variables such as the type of official, the frequency of contact with the public officials, the gender of the person in contact with the public official and other social and demographic characteristics, the coefficient on the gender of the public official is significant and positive. This suggests that the gender of the public official has an influential role in determining the likelihood of paying a bribe. This result conforms that their is a gender-specific dimension to the mechanism of corruption in dealing with public services.

Compared to individuals who interacted with both male and female officials of a certain type within a year, those solely interacting with male officials are 1.92 times more likely to have paid a bribe, while those exclusively dealing with female officials are 1.19 times more likely. This surprising finding, even when accounting for contact frequency, suggests that those interacting with both genders are the least likely to pay bribes. Notably, individuals dealing exclusively with male officials are 1.6 (1.92/1.19) times more likely to pay bribes than those exclusively interacting with female officials, holding other variables constant. It must be mentioned here that these results could be affected by the high survey weights in GIPSS 2021. As explained in Section 3, the model was run without the survey weights, and in that case only being in contact with a male official was significant. This highlights that especially being in contact with male officials is a driver of paying bribes.

5.3. Conclusion on corruption and gender of public official

In conclusion, this section has shown that also for the gender of public officials, there is still a significant difference in the likelihood that people paid a bribe based on whether the public official is male or female. It is difficult to pinpoint exactly what causes this, although the perceptions in Ghana are mainly that this is because men need to provide more for families, as well as that they are simply less afraid to request bribes.

Research with public officials on their reasons to request bribes would be extremely insightful to be able to develop accurate policies against bribery. However, the difficulty in retrieving such data should be acknowledge, as officials will most likely not want to answer questions on bribery truthfully. It is therefore an important task in the progress of research on bribery and especially the gender dimensions, to be able to find novel ways to really dig deeper into the thought processes of officials when requesting bribes.

8.1. Methodological recommendations

The analysis in this paper has only been possible because of the successful design and implementation of the 2021 Ghana Integrity of Public Services Survey (GIPSS). The survey methodology was designed to be a large-scale household survey representative of Ghana’s 16 administrative regions. It targeted 15,000 interviews and had a response rate of 99 percent. The sampling methodology adopted was three-stage stratified clustered sampling. The Primary Sampling Unit (PSU) was the Enumeration Areas (EAs). Using the household register from the latest population census in 2021, 10 households were selected at random from each EA.

The survey methodology followed closely the standard methodology adopted by many other countries as per the guidelines and standards of the UNODC. However, new to this survey was the inclusion of a gender specific section, which captured differences between men and women with respect to paying and receiving bribes, as well as gender specific perceptions of corruption. The inclusion of this section was very successful and yielded a lot of interesting results, most of which have been discussed in this paper.

In particular, the quality of the data collected during GIPSS is notably high. This has been accredited to two key aspects of the data collection process. Namely, the development of a real time monitoring dashboard which accurately tracked metrics such as the GPS location of interviews, the length of interviews, and number of interviews completed by each interviewer per day. These metrics were monitored daily and any issues addressed promptly.

Furthermore, training of participants (interviewers, supervisors, and facilitators) was thorough. Additional refresher training was also conducted in which everyone involved in the survey was trained again, with a specific focus on the main errors found in the preliminary data. The supervisors on the survey also ensured that any issues were quickly resolved. For example, due to sensitivity of the topic, some household members were reluctant to provide information, but the availability of monitors to explain the survey objectives, build rapport, and involve opinion leaders in the community eased the collection of data and ensured the 99 percent response rate.

Despite the overall success of the GIPSS there were also some challenges. Below, ten recommendations to improve the collection of corruption data and future research into corruption are suggested:

The novel inclusion of the gender dimension into the survey proved to be very insightful and yielded important results. Therefore, it is strongly recommended to include a gender dimension to all future corruption surveys. Moreover, future research could focus on whether greater gender equality in public offices matters for the incidence of corruption.

8.2. Policy recommendations

Based on the empirical evidence of the gender dimensions of bribery in Ghana, the following five policy recommendations are proposed:

Prioritise the successful implementation of the National Anti-Corruption Action Plan (NACAP) as well as national gender policies that aim to promote gender equality, equity and women’s empowerment at all levels and sectors. Putting a combined focus on anti-corruption as well as gender policies will be instrumental to reduce bribery.