Frequently Asked Questions
It is preferable to simply instruct the interviewer to fill in a zero or “no” when s/he already knows or can plainly see that the household does not possess or eat a particular item. An interviewer should be permitted to skip over questions when the answer is obvious, such as the possession of vacuum cleaners in mud huts, whether people ate only cassava in the last 7 days when their house is full of TVs, stereos and the like, or when members of the household serve the interviewer snacks. Some of the questions applied in very poor regions (ownership of a motorbike, a metal roof, etc.) will be out of place simply because the people are so much poorer than others in the country. There is no need to ask a question if the answer is obvious. It can be embarrassing for both the interviewer and interviewee to ask about something that poor people obviously do not have.
It is important to remember that we are not measuring relative poverty; we are measuring absolute poverty. Measuring relative poverty would entail comparing the households in one village to each other, but measuring absolute poverty is comparing the households in one village to households around the entire country.
There is no such thing as 'sampling for the PAT' per se. Sampling is a statistical technique used to find out information about a large group by only assessing a smaller subset of that group. There are many different ways to conduct sampling; several of them can be used to implement the PAT.
Chapter 3 of the PAT Implementation Manual (available in English, Spanish & French) discusses several sampling methodologies in detail. You can also complete the module on sampling in the online PAT training course. The course is free of charge and you can complete any sections of it individually. You will have to create an account to login; instructions are on the course homepage.
Yes! Pre-testing serves several important purposes. It helps you test your survey translations and provides valuable practice for your survey implementers. It can also be used to verify client information (addresses, names, etc).
The best way to maintain the accuracy of the assessment tools during translation is to ensure that the translation is as similar to the original version as possible. As described in the Tool Implementation Manual, the best way to accomplish this is to translate the tool into the desired language, and then translate it back into English. This process is an excellent way to ensure that meaning is not lost in the translation process. Translation is best accomplished before any interviewing from the final sample of clients begins; pre-testing of the questionnaire and implementation process on a small sample of clients is recommended. It would compromise the accuracy of the assessment tools to simply translate the English versions of the tools “on the spot” during an interview with a client, for any language or dialect.
A number of PAT implementers have determined, after using the PAT for the first time, that they should have spent more time refining survey translations into local languages. Here are several specific challenges implementers have noted:
-- Implementers found that specific words or phrases use in the survey were inappropriate for the context in which they were administering the survey.
-- Implementers found that the intention of the question was sometimes unclear. (Example: When asked, "Do you own a bed?", respondents to one survey were unclear what constituted a 'bed' (something on a bed frame, a rolled up mat, a 'local' bed stuffed with straw, etc. In this case, the word used to represent 'bed' was not specific enough and a different word should have been used.)
-- Because we did not come to a consensus on the translation with all interviewers, several made changes of their own while administering the survey. This led to inconsistent results.
-- In some cases, surveys were not translated out of English in writing; interviewers were fluent in English and the local language and interpreted the questions while asking the respondents. Even though some interviewers preferred it this way (because they were only used to speaking and not reading their local language), it was determined that the surveys were interpreted in very different ways, leading to inconsistent results.
CSPro and Epi Info were chosen for PAT analysis because of their user-friendliness and ability to conduct multiple analyses. In May 2013, a Data Analysis Manual was released to help PAT implementers who wish to analyze PAT data in other software.