Survey Universe and Methods

Survey Universe

Our survey universe was pulled from the Missouri Municipal League’s listing of Missouri municipalities (excluding only St. Louis and Kansas City). We administered the survey in phases, beginning with the areas surrounding Kansas City and St. Louis, and then expanding into the rest of the state. This allowed us to insure that an appropriate mix of urban and rural municipalities were represented while keeping our demographic data questions to a minimum (for instance we didn’t have to ask what region in the state the respondent was from). This was meant to maintain the respondent’s sense of anonymity throughout the survey-taking process. The respondents were a variety of public officials and employees across a broad range of municipal sizes.

chart05
chart06


Survey Methods

An initial decision was made to administer the survey using a widely-recognized, publicly-available online survey service – Survey Monkey – which the respondents may have encountered before and with which they were perhaps familiar and more comfortable than other methods. In contacting the municipalities we began with a phone call, the script for which emphasized the anonymity of the survey. If the contact indicated a willingness to take part, we forwarded an email that again emphasized anonymity and contained a link to the survey.

We hoped that the anonymity of the survey and use of a popular service would reduce the social desirability bias, or fear of reporting situations that may be unethical. However, it is likely that as the moral condemnation attached to a particular infraction increases, the more likely it is that the respondent will not feel comfortable honestly reporting on the frequency with which she has observed such actions. Actions that are directly criminal, and understood as intentional corruption (such as embezzlement and theft) are therefore theoretically much more subject to that bias than actions that may be unethical, but not intentionally criminal or corrupt (such as carelessness with confidential information, or creating a hostile work environment). Intentionally criminal or corrupt actions are also presumably more difficult to observe since the actor will likely take precautions to avoid observation. Though our survey produces interesting results on the state of traditional corruption in local government, they should be viewed in light of these considerations.

Our analysis consists of comparing the proportion of survey participant subgroups who observe each of the varieties of ethics violations. The main subgroups of interest are based on our main independent variables:

  1. The population size of the municipality in which a participant works;
  2. Whether a participant’s agency has an ethics code; and
  3. The level of emphasis on ethics training a participant reports in their agency.

The statistical methods we employ identify whether there are statistically significant differences in ethics violations across population size, institutionalized ethics codes, and levels of training emphasis. We acknowledge that the data collected on ethics violations are generally skewed to the right, which would skew reported means, and are measured at the ordinal level. As such, difference-of-median tests were used to account for non-normality and crude level of measurement. We use spearman correlations to measure the magnitude of the association between ethics violation category and the independent variable.