In March 1997, the University of Florida's Bureau of Economic and Business Research (BEBR) conducted a telephone survey of 1,002 Floridians at the request of the Florida Department of Corrections (DC). In June they conducted an almost identical survey of news media reporters in Florida. The purpose of these surveys was to gather baseline data on the general public and media's knowledge and image of the DC, as required in the DC's Agency Strategic Plan, Issue 6, whose goal is to "keep staff and the public well-informed of agency issues, policies and practices." Further plans include repeating the surveys annually to determine whether subsequent efforts at educating the public and news media about correctional issues, policies and practices have been successful.
Quotes were solicited from all universities in the State University System to conduct the general public survey. Of the five who responded, the University of Florida's BEBR was selected based on their extensive experience, lowest cost ($3,276), and their use of the same statistical software (SAS) as the department. BEBR also conducted the news media survey at a cost of $3,078 for 366 completed surveys, and the DC staff survey for $5,400 for 699 completed surveys.
In March, the BEBR added 44 questions for the Florida Department of Corrections to its monthly survey of Consumer Attitudes of Floridians over 18 years of age. A total of 1,002 respondents were surveyed by telephone and 936 of those individuals completed our entire survey. The calling areas were stratified by four regions: Tampa, Orlando, southeast Florida, and the rest of the state. The survey was conducted by BEBR by telephone between March 1 and March 31, 1997. Nearly 95 percent of Florida households have a telephone. Twenty-three percent of those surveyed had unlisted phone numbers. The cooperation rate ((completes)/(completes + refusals))for March was 54%. The disposition of responses is as follows: 936 completes, 814 refusals, 761 ineligible (underage), 646 non-working, 506 no answer, and 156 incomplete.
In June 1997, the BEBR attempted to survey a corrections or courthouse reporter from all 708 media outlets listed in the 1997 Florida News Media Directory. The media outlets that cater to special subjects, such as shopping or cars, were not surveyed. The outlets include daily and weekly newspapers, television and radio stations and news services. After four attempts to interview the corrections/courthouse reporter, the BEBR asked for a general reporter or the news director. BEBR increased the number of calls before finalizing a case from their standard 10 to 15, and ultimately obtained 366 completed surveys. As in the general public survey, all interviews were computerized. The cooperation rate among outlets that had reporters was 93%. The disposition is as follows: 366 completes, 28 refusals, 51 not accessible, 230 no news or no news reporters, 26 non-working, and seven had no answer.
In August and September 1997, the BEBR surveyed 699 Department of Corrections staff throughout Florida. These staff members were selected randomly from a list of all DC employees, and included 398 non-security and 301 security staff from regions, institutions, probation and parole offices, central office and all other facilities. All interviews were computerized. The cooperation rate was 93%. The disposition is as follows: 699 completes, 50 refusals, 211 not accessible (often those working late shift in institutions), four not applicable, 37 non working numbers or employee moved, three no answer.
It should be noted that this survey was conducted during the same month (March 1997) that several hundred inmates had to be released from prison early as a result of a Supreme Court decision (Lynce decision), which received extensive media attention for several days. A t-test was conducted to determine whether responses before the inmate release differed substantially from responses after the inmate release. (A t-test is a method used to measure the difference between two averages to determine if there is a statistically significant difference.) The only statistically significant difference in the variables reviewed was in the final question regarding an overall evaluation of the Florida prison system. While those who answered "fair," "good" and "excellent" to that question were fairly evenly distributed before and after the Lynce decision made the news, 71% of those who responded "poor" to that question did so after the Lynce decision was announced. Also during March 1997, there was a problem with an execution, but no t-test was conducted concerning this issue and no questions pertaining to execution were asked in the survey.
Some items vary significantly by region; therefore, the general public data have been weighted by share of households to correct for oversampling. The weighted responses were then rounded to the nearest whole number. Telephone numbers were generated by Random Digit Dialing, where random numbers are attached to working telephone banks (area code -prefix-suffix combinations that are known to be residential areas). All interviews were computerized and responses were edited upon input using preset checks and ranges. Those listed as "Cases Not Applicable" in each question consist of those who did not complete the survey, and those who answered "don't know" or "not applicable" to the question. Those listed as "Cases Not Applicable" were not used in calculating valid percentages.
Weighting was not necessary for the news media survey because every potential respondent was contacted.
The DC staff survey was not weighted.
Apart from the potential error introduced by response bias (those who refuse to participate), all surveys have some error due to the fact that all persons in the target population were not contacted. When there is variation in responses to a question, which is common, the possibility exists that if the survey were conducted again, different results would be obtained. This is the margin of error. The margin of error can be estimated based on the variance of responses and the number of interviews. The statewide estimates for this survey have a margin of error no more than three percent. The actual margin of error for each estimate depends on the actual distribution of responses.
Since the entire population of media outlets listed in the 1997 Florida News Media Directory was contacted, there is no margin of error for this survey.
The maximum margin of error for the DC staff survey is plus or minus 3.7 percent. The margin of error is smaller for responses with lower variance, which includes questions that require more than a yes/no response.