Blog Question: What are appropriate purposes for surveys, how are subjects selected, how is data collected and analyzed, and what kinds of generalizations are possible?
Surveys are descriptive according to Lauer and Asher (L & A), surveys are often very analytical and potentially quantitative. Surveys are well suited for complex and complicated research efforts that must be executed within a limited resource situation. Surveys reduce the target population to a manageable size with reasonable results. While surveys can be expensive, they may be cheaper that full experimental research. However, surveys are not just descriptive as L &A suggest. It is true that surveys are a wonderful tool for inferring about the descriptive variables of a population, but survey analysis could be considered (if we think back to Morgan) to be a correlational. Not only can a survey’s goals to describe or inform variables, but it can show relationships between variable depending on the questions and design of the survey. In particular, people’s attitudes are easier (and I use “easily” loosely) to correlate depending on their answers to various answers. For example, individual income and commute times normally have a relationship within urban research.
Subject selection is the crux of any survey. Essentially, the subject selection is a large influence on validity and if the sample subject selection is off then the validity is not there. As a result, there are a host of various sampling approaches, random, systematic random, quota, cluster, and stratified sampling. The type of sampling approach depends on the researcher’s goals and desired data. Random sampling works well for an amorphous population. Stratified sampling works best when the researcher is dividing or distinguishing between populations and quota sampling works well with representing certain subgroups within a larger sample. The analysis of the survey results, (for me) is normally a statistical analysis. The analysis must determine whether the sample is representative, statistically significant sample size, and then identify the significant variables and relationships. Within planning, identifying or describing variables is important, but correlational and eventually causal analysis of variables is demanded (although causal assertions can be technically inaccurate, city leaders normally demand “sure things”). Finally, surveys benefit from other methods and research. Adding a minor case study or quantitative analysis can increase the potency of the survey data.
Unlike case studies, a properly designed survey intends to draw inference about a population based on the sample or samples. If the survey is inappropriately designed or does not sample enough to represent the population, the researcher cannot generalize. The research must remain careful not to assert or claim things that the survey data does not support. The wording and execution of the survey does affect the researcher claim. Asking, “will you support transit” is not the same as a willingness to actually ride transit. In addition there is bias and cognitive dissonance that may not surface in the survey. An famous/infamous national survey about gasoline price asked “if gas cost $3, would you cease driving and rely solely on transit” … well 70 percent of the respondents said they would switch to transit; however, with the summer peak of gas prices, this claim by the survey respondents really did not happen. Another example with Clemson ridesharing, only 30 percent were not interested in a carpool program, but this number jumped to over 40 percent when asked to have their name entered into a database of potential users. The moral of this story: just because the survey responds in a particularly to a prompt, this does not mean the respondent will actually act this way when confronted in real life.