There are lies, damn lies and statistics/surveys/IQ-tests.
One of the best examples of surveys being bunk in general is the US election time surveys. They are almost always exclusively wrong. They might be right in calling majority on a candidate who wins, but most of the time they have been horribly wrong in closeness of the race.
Other things to consider in these types of surveys:
- Error margin and sample size.
- Asking the same question, only translating it to a native language without considering the context or meaning of the same question to local population/culture.
- Taking in to account the concept of racism on local population. Some people don't even understand what racism is, certainly true in bangladesh.
- Racism awareness usually comes as a result of struggle and long standing local/country history. If the country or culture didn't face any such issue, they are less likely to understand the concept.
- Quality of education can fill in the blanks when there is lack of first hand experience. Poor or developing countries usually have low education standard.
- Even the definition of racism can differ from culture to culture. Depends on who says it, how he says it or in what context he says it, they definition of "nigger" can swing from person to person.
- In case of Australia, which has a negative connotation of being a racist country (so does japan), their awareness of what constitutes as racism might also make them less likely to participate in a survey with the right information, even if the survey is anonymous.
- Cognitive Dissonance.
Last but not least, the example of ~80% of BD people wanting a sharia law. While I don't remember the exact details of said survey, but I think I remember reading the sample size being 1000 people. Yes questions asked to 1000 people in a country of 150 million decides how rest of the country feels about sharia law. Not to mention, there is a negative connotation among religious people (which is most of BD) about saying or even thinking anything even remotely negative towards religion (islam shommonde kharap chinta korleo ghuna hobe), no surprise that people are going to say things in favor of religion, without taking in to account any implications. Also did the people who participated in said survey understand the concept of anonymous survey, most likely not.
So you have to take all these things in to consideration. Survey are not black and white, and doesn't work in all kind of questions. They usually work in certain very strict and definite questions. Questions that are up for interpretation and generalized are useless surveys.
Thats not to say that most people in Bangladesh are not racist, they are to western standard, thats most likely because of lack of education and not understanding the concept of racism.