1.3. The reliability of responses
The views of respondents making a large number of assessments
The survey sought to capture the knowledge of respondents regarding their experience of companies from up to 30 countries. However, as some respondents provided answers about companies from a very large number of countries (2.6 percent of the respondents assessed all 30 countries), it can be argued that these respondents may not have precise experience of how companies from so many countries do business. Indeed, it could be that these respondents misunderstood the question, believing that they were being asked to give their impression of all countries rather than their experience of just those with which they were familiar. The histogram below, in figure 2, shows the frequency distribution of the number of assessments made by all respondents to the survey, between 1 and 30.1
- Figure 2: Frequency distribution of the number of assessments (countries) made by each respondent
Taking this histogram into account, it is possible to assess the possible impact of this perception bias by analysing the results excluding the answers of any respondent who provided answers about more than 20 countries. 2 This analysis produces results which show an extremely high correlation (0.99) with the BPI 2006, but with some changes in the middle of the ranking. Excluding respondents who had made more than 20 assessments reduced the sample size by 30 percent (number of country assessments) and caused the standard deviations and confidence intervals to increase. This reduces the precision of the results. Therefore, it can be concluded that excluding respondents who had answered for more than 20 countries would not significantly change the results of the BPI 2006 and indeed would affect the statistical strength of the sample. The full sample was therefore used for the BPI 2006.
The views of respondents from foreign and locally owned companies
As the sample contains both foreign and locally owned companies, and the questions used in the survey assess bribery and undocumented payments in international business, it could be argued that the different types of respondents have different experiences of bribery. To assess this potential bias, companies which are at least 50 percent owned by the foreign private sector are defined as ‘foreign owned’, and companies with at least 50 percent ownership by the local private sector or the government are defined as ‘locally owned’. The results of this analysis show that foreign owned companies tend to have experienced a lower incidence of bribery than their locally owned counterparts. Figure 3 demonstrates that this distinction is particularly pronounced for countries that rank in the top 15 of the BPI 2006. At this point it is not possible to establish whether this difference is associated with bias or with experience, especially as there is less divergence towards the bottom of the ranking. This question should be subject to further research. The full results of this analysis are shown in table 3, Annex 2.
- Figure 3: How foreign and locally-owned companies view foreign firms.
The views of respondents from small and large locally owned companies
When taking just locally owned companies into consideration, the BPI 2006 results can be further disaggregated to examine the effect of company size on respondents’ knowledge of international bribery. Companies with 100 employees or less are defined as ‘small’, those with 101 to 500 employees are defined as ‘medium’, and companies with more than 500 employees are defined as ‘large’.3 The results show that small locally owned companies tend to report more instances of corruption than large ones, and medium sized companies tend to report less instances of corruption than small companies, but more than large companies. However, as the difference between the three sizes of company appears to be consistent across the ranking, and it is not possible to assess which of the three is making a more accurate assessment, isolating potential biases related to company size at the local level does not seem to improve the quality of the conclusions. Furthermore, the responses of large locally owned companies are comparable to those of foreign owned companies as reported earlier in this section. Therefore a similar reasoning applies to this category, namely that it is difficult to establish if the respondents’ answers are biased or based on experience. The full results of this analysis are shown in table 3, Annex 2.
- Figure 4: How small, medium and large locally owned companies view foreign firms
1 3,198 of the 11,232 respondents surveyed (28 percent) did not offer an assessment on any country regarding the propensity of their firms to bribe abroad. This rate is not unusual for questions related more to experience than perception. These non-responses could reflect a lack of knowledge or an unwillingness to answer. The BPI 2006 was therefore calculated using the scores given by the 8,034 respondents who did offer an assessment of companies from at least one country. For the sake of presentation, in this histogram the 3,198 respondents that made no assessments are not included.
2 Various scenarios were considered, including up to 15 responses or between 3 and 9 based on the histogram. For simplicity, we report here the results for the cut at up to 20 respondents, where there is an indicative break according to the histogram.
3 There is no universal definition of small, medium and large companies. For the purposes of this analysis, divisions of 100 and 500 employees were used. This matches several commonly used definitions of company size.