Comments on Prothom Alo Polls
While voting, in general, is probably the best way to sort out the best option out of many, it is not necessarily a fair way when more than one options are available to the voter. In an electronic voting, fortunately the option is often times limited to two, e.g., yes or no, which makes it a fair assessment as long as the system can be trusted (without bugs or predetermined outcome, which an intelligent programmer can sometimes code into the program) for reliability and accuracy. For a reliability of opinion, such votes must also come from random samples and not a particular biased group. The most important parameters, however, are sample size (based on population size), degree of confidence and margin of error the results would have.
So, e.g., if we have a population size of 1 million, and we are looking for a 95% confidence (i.e, a take a risk of 5% that I could be wrong with my assessment), then we shall need a sample size of 2400 voters to vote. The margin of error will be 2%. As the population size increases, the sample size requirement increases up to a point (being an exponential type function), plateauing out above 1 million. Thus, if we have a small population to begin with, the sample size requirement would be smaller.
With this background, if we are to study, e.g., the Prothom Alo results about the CHT problem, we can say that the the sample size is adequate for a test of 95% confidence. As to the result itself, the difference of 54% for yes vs. 42% for no (ignoring the third option - undecided) shows that the 2% margin of error will not have altered the general conclusion that the current government cannot solve the CHT problem satisfactorily.
There are, however, some catches with many electronic votings that we need to be acquainted with. Some systems allow a single person to cast multiple votes, i.e., one can hit the yes- or no-button as many times as one wishes, without being restricted to do so. Even if a restriction is imposed, it is often limited to a username or a machine from where the polling site was accessed. Thus, if a person has multiple usernames and/or multiple machines, he or she could cast his/her vote multiple times, thus affecting the outcome of voting.
Since I don't usually participate in such e-votings, I don't know how reliable the Prothom Alo e-voting was. We must also take into consideration the preference of people accessing the Prothom Alo site to begin with. Who typically are the readers of the site? Do they belong to a certain ideology, age group, demography, profession, etc.? Could some activist for a particular opinion lobby or influence others to vote for the same opinion as his or her?
Without all such info available, the results of an opinion poll are just a piece of info that may not be reliable.
So, e.g., if we have a population size of 1 million, and we are looking for a 95% confidence (i.e, a take a risk of 5% that I could be wrong with my assessment), then we shall need a sample size of 2400 voters to vote. The margin of error will be 2%. As the population size increases, the sample size requirement increases up to a point (being an exponential type function), plateauing out above 1 million. Thus, if we have a small population to begin with, the sample size requirement would be smaller.
With this background, if we are to study, e.g., the Prothom Alo results about the CHT problem, we can say that the the sample size is adequate for a test of 95% confidence. As to the result itself, the difference of 54% for yes vs. 42% for no (ignoring the third option - undecided) shows that the 2% margin of error will not have altered the general conclusion that the current government cannot solve the CHT problem satisfactorily.
There are, however, some catches with many electronic votings that we need to be acquainted with. Some systems allow a single person to cast multiple votes, i.e., one can hit the yes- or no-button as many times as one wishes, without being restricted to do so. Even if a restriction is imposed, it is often limited to a username or a machine from where the polling site was accessed. Thus, if a person has multiple usernames and/or multiple machines, he or she could cast his/her vote multiple times, thus affecting the outcome of voting.
Since I don't usually participate in such e-votings, I don't know how reliable the Prothom Alo e-voting was. We must also take into consideration the preference of people accessing the Prothom Alo site to begin with. Who typically are the readers of the site? Do they belong to a certain ideology, age group, demography, profession, etc.? Could some activist for a particular opinion lobby or influence others to vote for the same opinion as his or her?
Without all such info available, the results of an opinion poll are just a piece of info that may not be reliable.
Comments
Post a Comment