Question

The U.S. Bureau of Labor Statistics (BLS) wishes to estimate the fraction of individuals who belong to a labor union. They are considering two methods of sampling. For each method, explain if you think it will be subject to selection bias (Yes or No) and why, and if it will be subject to non-response bias (Yes or No) and why. Also state whether you think each method will result in too small (downward bias) or too large (upward bias) an estimate of the true average.

a)BLS will call a random sample of phone numbers from each town and city’s phone book during regular business hours (9 AM to 5 PM) and ask whether or not the person who answers belongs to a union.

b)BLS will survey people at the entrance of each factory in the country on weekday mornings and ask them if they belong to a union.

c)Now propose a sampling method of your own that you think will generate a more accurate national average. Explain why you think it is better.

Answer 1

Selection Bias

In this kind of sampling, the population taken for analysis will include some samples who are not a part of the population. It means certain unnecessary samples may enter the population for analysis. One example is undercoverage. In this kind of bias, the members contained in the population will be non-representative samples. Mainly, this selection bias occurs when the selection of samples are influenced by exposure status.

Non-response Bias

In this kind of sampling, samples taken will not be much co-operative. Individuals take for the survey may not be willing to be a part of the same. They will neglect to respond. Mainly, it happens when the medium of survey will not match to the group of individuals who are taking part of the survey. One example is when an ordinary mail is send take survey of youth, they may not respond to the same as the medium of survey is not interested for them.Hence, the rate of response will be very low.

Upward bias

The measure of survey will be overestimated, even more than what the survey intend to explain. Hence, this bias is also called upward-bias. This will be effected because of the omitting of some variables.

Downward bias

The measure of survey will be underestimated, even less than what the survey intend to explain. Hence, this bias is also called underestimated bias. This will be effected because of the impact of some explanatory variables.

Omitted variable bias.

Both the upward and downward bias are part of omitted variable bias. Omitted variable bias means when an estimation is considered, any key factor to be included in the survey was omitted due to insufficient data of that variable.

• Now, let us review each methods of sampling:-

a.)

The U.S. Bureau of Labor Statistics will call a random sample of phone numbers from each town and city’s phone book during regular business hours (9 AM to 5 PM) and ask whether or not the person who answers belongs to a union.

This method will be subject to non-response bias as elder people included in the survey may not completely respond to such a telephonic survey. Also, this method will result in downward bias, which is too small because the U.S. Bureau of Labor Statistics are calling the people with the help of phone number that are registered in the city phone book. This kind of survey may omit certain variables due to unavailability as every individual in the city may not have a landline telephone. Another problem comes under the same is BLS is calling for survey during the business hours. Several people will go for work this time and cannot participate in the survey. By knowing this fact, BLS is still conducting the survey. Hence, several people who may be the part of Labor Union will be omitted from the survey though they should be included in the survey. Hence, this method will be subject to selection bias.

b)

BLS will survey people at the entrance of each factory in the country on weekday mornings and ask them if they belong to a union.

This method will be subject to a response bias because people may not be reluctant to participate in a survey taken face-to-face. Also, this method will result in upward bias, which is too high because the U.S. Bureau of Labor Statistics are taking the survey of all labors who are working in every factory in the country. All of the workers will not be member of the labor union. Hence, the sample taken in the population for the survey is too big. Hence, this method will be subject to selection bias.

c.)

Propose a sampling method of your own that you think will generate a more accurate national average.

Stratified sampling is the best method I opt for this kind of a survey. Stratified sampling ensures the statistical measures of a particular population. Here, members of the labor union. Before taking the sample, population has to be divided on the basis of their characteristic. Then, sample should be taken from the population. This ensures that the needed representatives samples are included in the population.