Survey Methods

How the Survey Was Conducted

The survey was conducted by the Marist College Institute for Public Opinion and funded by the Dyson Foundation.

4,443 adults 18 years of age or older who live in the Mid-Hudson Valley in New York were interviewed in proportion to the adult population in seven counties: Columbia, Dutchess, Greene, Orange, Putnam, Sullivan, and Ulster. The survey was designed to collect information about the Mid-Hudson Valley as a region and to allow for county level analysis as well. Representative samples were drawn from each county based on the population distribution within that county. In order to analyze the opinions of residents of the Mid-Hudson Valley, the seven county samples were combined and weighted to reflect the population distribution of the entire region. Census comparisons to the 2012 survey results are displayed at the end of this report.

This is the third time over the past decade this project has been undertaken. Many of the questions in this survey were asked of Mid-Hudson Valley residents in each of the previous two studies. As a result, comparisons may be made over time for 2007 and 2002. Residents of Sullivan County were first included in the 2007 study of the region.

The goal of a scientifically designed survey sample is to be representative of the population that is being surveyed. The results obtained from a scientific probability survey are not just answers from those individuals who responded but more importantly, because of the design and methods by which the data is collected, can be used to generalize to the population as a whole. For this survey, the results are an estimate of what would have been obtained, within a certain range, if all adults 18 years of age or older in the Mid-Hudson region were interviewed.

How to interpret the numbers

Sampling error is the difference between the responses to the survey if all adults 18 years of age or older who live in the Mid-Hudson Valley had been interviewed and the actual survey results. Sampling error is primarily based upon the number of interviews in the survey sample. The sampling error may be interpreted as indicating the probability (95 times out of 100) within which the results of repeated samplings, in the same time period, assuming the same sampling procedures, could be expected to fall within a certain range.

The margin of error for the survey results of Mid-Hudson residents is ±1.5% for percentages near 50% at a confidence level of 95%. The sampling error diminishes slightly for questions whose results are at the extremes and the sampling error increases as the number of interviews for a particular group or sub-group within the sample declines.

For example, 51% of Mid-Hudson Valley residents surveyed think their community spends too little money on public transportation. The margin of error of ±1.5% means that we may conclude there is a high probability, 95 times out of 100, that the average results for this question of repeated samplings in the region will fall between 49.5% and 52.5%. All results are rounded to the next whole number.

In evaluating the information detailed in this report, several points should be noted. First, the analysis often focuses on differences among subgroups rather than commonalities. Therefore, the charts depicted in the report do not include all factors but rather those where differences are evident. Additional tables may be found at ManyVoicesOneValley.org and may be referenced for further detail among subgroups and other key segmenting variables.

Second, some questions in the survey were only asked of a subgroup of those interviewed depending on an answer to a previous question. For instance, residents were asked if they were currently employed. If they were not, they were not asked subsequent questions about their job such as commute time. Questions that were not asked of all residents are identified.

Finally, totals in tables may not add to 100% due to rounding.

Methodology: sample design

A dual frame, stratified random digit dial (RDD) probability design was used to draw the telephone numbers for the survey. RDD ensures representation of both listed and unlisted telephone numbers. Telephone numbers were selected based upon two separate frames or lists of telephone exchanges, one for landline and one for cell phone numbers. Telephone numbers were selected based upon a list of telephone exchanges from throughout the seven counties. The exchanges were selected to ensure that each county was represented in proportion to its population in accordance with a probability sample design that gives all landline telephone numbers, listed and unlisted, and cell phone numbers a known chance of being included. Results are based on both landline and cell phone interviews of adults 18 years of age and older. The landline and cell phone samples were combined. Nineteen percent of the completed interviews were obtained from the cell phone sampling frame.

In order to participate in the survey a household or cell phone owner had to have a residence located within one of the seven counties and have an adult resident that was 18 years of age or older. For the landline frame, a member within each household was then selected to be interviewed through the use of a random household selection method. In order to be interviewed, a respondent needed to be at least 18 years of age or older and be a resident of the Mid-Hudson Valley.

Methodology: data collection

The questionnaire and the telephone sample were programmed for computer assisted telephone interviewing (CATI). Interviewing was conducted from a centralized telephone facility at the Marist College Institute for Public Opinion using trained interviewers who were specifically briefed on this study. The interviews were conducted in either English or Spanish.

There were a total of 294 interviewers and supervisors who worked on this study. On average, the supervisor to interviewer ratio was seven to one. Polling supervisors regularly monitored, evaluated, and provided feedback to the interviewing staff throughout the data collection period. The average length of the survey was twenty-two minutes.

One hundred and ninety seven pretest interviews for the survey were administered by telephone on December 13th, 2011. As a result of the pretest, the questionnaire was updated and revised. Although there were limited changes made to the questionnaire, pretest interviews were not included as part of the final dataset.

Interviews were conducted February 2nd through March 14th, 2012. Interviewers contacted households between 5:30 p.m. and 9:15 p.m. Callbacks were also conducted between 10:00 a.m. and 4:00 p.m. on weekdays. Up to 8 attempts to contact a household were made. Callbacks were staggered over times of day and days of the week to maximize the chances of making contact with a potential respondent. Suspended interviews and refusals were re-contacted at least once in order to attempt to convert them to a completed interview. This included callbacks to initial unsuccessful contacts, refusal conversion, and scheduled appointments. A toll free number was used so that respondents could call the survey center at their convenience to complete an interview.

Information from survey participants is both confidential and anonymous. Personal identifying information is removed from files after the integrity of the data is verified.

Comparisons of 2012 Survey Results to 2010 Census