You are here

Methodology

Methodology

BTS and MARAD developed a four-page, 15-item questionnaire (available upon request) to collect information from a sample of merchant mariners. The first mailing of the questionnaire occurred on March 26th, 2001. A follow-up mailing was conducted on April 27th for nonrespondents whose questionnaires had not been returned as undeliverable by the postal service. Beginning May 15th, a telephone follow-up was conducted with a sample of mariners who had not yet responded to the mail survey. Data collection was closed on July 9th, 2001, ending a 15-week data collection period.

Sample Selection

The U.S. Coast Guard is the lead federal agency for regulating, licensing, and documenting professional merchant mariners. The U.S. Coast Guard’s Merchant Mariner Licensing and Documentation (MMLD) system provides information on the qualifications and reported seafaring employment for both actively sailing and inactive mariners.

At the time of the survey, the MMLD contained information on over 210,000 individuals. From that file, a target population was identified of all mariners with documents appropriate for deep-sea employment as an officer (Unlimited License Holders) or other crewmember (Other Mariners) who had completed a document transaction within the last 6 years.

The 6-year time frame was selected because all mariners are required to renew licenses and documents on a 5-year cycle. Extending the time frame to 6 years allowed inclusion of mariners whose documents were less than 12 months out of date on the assumption that some proportion of these mariners would renew before the 12-month window for renewal had elapsed.

Based on the limited response period and the emphasis on reaching mariners who would be most readily available to crew large ocean-going vessels in the event of a national emergency, a small number of individuals with military addresses and/or foreign addresses were removed from the target population.

The final target population contained a total of 104,170 mariners—20,157 Unlimited License (UL) Holders and 84,013 Other Mariners. A sample of 10,000 mariners was randomly selected to participate in the survey. UL Holders were over-sampled to ensure sufficient numbers from this group to maintain acceptable confidence intervals for population and sub-group (e.g., rank or specialty) estimates. The final sample included 4,550 UL Holders and 5,450 Other Mariners.

Reliability of the Estimates

The findings summarized in this report are estimates derived from a sample survey. There are two major components of error in a sample survey—sampling and nonsampling error.

Sampling Error. Sampling error occurs because findings are based on a sample, rather than on the entire population. The total respondent pool for the mariner survey is 4,046 for an estimated sampling error of about ±2% at the 95% confidence level. This means that if a comparably sized series of additional samples was drawn from the current mariner population, 95% of the time results from those samples would fall within a range of ±2% of the results reported for the current sample. Sampling errors are larger for sample subgroups, such as mariners with deep-sea experience (±3%) because the respondent pool is smaller.

Nonsampling Error. Estimates are subject to various errors during the survey process, such as data collection, response coding, transcription, and data editing errors. These errors would also occur if a complete census was conducted under the same conditions as the sample survey. Explicit measures of the effects of these errors are not available. However, stringent quality control procedures were followed during data entry, and the questionnaire was reviewed and pretested in an effort to minimize nonsampling errors associated with data entry and questionnaire design.

Nonresponse error is afunction of both thenonresponse rate and thedifferences, if any, between respondents andnonrespondents. In an effort to quantify nonresponse error, a sample of nonrespondents was selected to receive the survey using a telephone interview. Of the 2,000 nonrespondents selected, 1,282 were located and 765 completed the survey. Because the mail survey had to be modified to fit Computer Assisted Telephone Interviewing (CATI) requirements, additional analysis is needed to assess the comparability of mail and telephone results. A report summarizing the telephone survey results will be available by September 30, 2001.

Response Rates

Final response rates for the mail survey were calculated using the American Association for Public Opinion Research guidelines (known as RR3) and are reported in Table 1.

UL Holders (55%) were much more likely than Other Mariners (29%) to complete a survey questionnaire. One reason for the lower response rate for Other Mariners was the much higher incidence of incorrect mailing addresses for this group. Over 1,000 Other Mariners (19%) did not receive a survey because their address information was incorrect compared to 479 incorrect addresses for UL Holders (11%).