Poll error not acceptable
Re: "Error margins okay: Abac", March 4,
Abac's excuses for being wrong were meek and insulting to our intelligence. It predicted that Pongsapat would win with 45.9 per cent over 34.1 per cent for his opponent. That margin of error of 11.8 per cent is not okay and it is irresponsible to say so. The margin is wider than the generally accepted statistical error of plus or minus 3 per cent. Only Nida Poll was almost spot-on. Dusit Poll chose not to conduct an exit poll after criticism over its neutrality.
Most of Thailand's pollsters thought they were the gurus in naming a winner instead of sticking to the statistics. Respectable pollsters stick to figures. The figures or graphs are always annotated with that "margin of error" caveat, which speaks volume about the degree of reliability. If the Thai pollsters had used that notation and refrained from naming names, they would not have felt embarrassed yesterday. Only Abac's error is beyond that standard statistical deviation.
In my view, pollsters talk too much on their surveys instead of sticking to statistics and what they're trained to do.