Appendix A

National Drug Threat Survey 2003 Methodology

The National Drug Intelligence Center (NDIC) National Drug Threat Survey 2003 (NDTS 2003) was administered to a probability-based sample of state and local law enforcement agencies. The sample was designed to provide representative data at national, regional, and state levels for use in the National Drug Threat Assessment 2004. The previous NDTS 2002 sample was designed to provide representative data at the national and regional levels only. The availability of state-level representative data not only increases the precision of the data used in this year's National Drug Threat Assessment but also expands the application of NDTS 2003 results to NDIC's state and regional threat assessments.

Survey Instrument

The NDTS 2003 questionnaire (OMB Number 1105-0071) was designed by NDIC. A thorough review of data and response patterns from previous versions of the NDTS was conducted to improve the accuracy of information obtained from respondents. Responding law enforcement agencies were asked to identify the drug that poses the greatest threat, that most contributes to violent crime, and that most contributes to property crime in their areas. Agencies also were asked to rate the overall level of availability (on a scale of low, moderate, or high) of powder cocaine, crack cocaine, heroin, methamphetamine, marijuana, MDMA (ecstasy), and other dangerous drugs in their area. The survey included an item designed to solicit information on the level of involvement of street gangs and outlaw motorcycle gangs in the distribution of drugs in general and of specific drugs. Other items in the questionnaire asked respondents to indicate the types of heroin available, predominant type of heroin, presence of crack cocaine conversion sites, presence of MDMA production laboratories, level of methamphetamine production, and nature of cannabis cultivation in their areas. Respondents also were asked to indicate which chemicals are diverted in or from their areas for the production of illicit drugs and which pharmaceuticals are commonly diverted or illicitly used in their areas.

  To Top     To Contents

Sample Design

The 2000 Census of State and Local Law Enforcement Agencies conducted by the U.S. Bureau of Justice Statistics was the basis for determining a sample frame from which to select law enforcement agencies to be surveyed for the NDTS 2003. After careful review of the more than 17,000 law enforcement agencies in the 2000 Census of State and Local Law Enforcement Agencies, a final sample frame of 7,930 state and local law enforcement agencies with drug law enforcement responsibilities was created. Municipal police departments from every state, including regional and county police departments with 10 or more sworn full time equivalent (FTE) employees, were retained for the sampling frame. County sheriff's offices with 10 or more sworn FTE employees were also retained for the sampling frame except those in six states where county sheriff's offices do not have drug law enforcement responsibilities. In the rest of the country, sheriff's offices were excluded if they did not indicate on the 2000 Census of State and Local Law Enforcement Agencies that they enforce drug laws. Campus police departments, constables, and special police agencies were excluded since most of these agencies, too, have limited or no drug investigation responsibilities. Tribal police departments, whose jurisdictions fall under federal authority, also were eliminated. State drug investigative agencies not in the 2000 Census of State and Local Law Enforcement Agencies were added to the sampling universe.

The sample frame of 7,930 state and local law enforcement agencies was stratified (see Table A1) to include the following specific groups of state and local law enforcement agencies to ensure a thorough analysis of the domestic drug situation:

• Municipal police departments and county sheriff's offices with 75 or more sworn FTE employees as reported in the 2000 Census of State and Local Law Enforcement Agencies were selected with certainty (stratum 97).
   
• State police and state-level investigative agencies were selected with certainty to provide information on the drug threat situation from a state perspective. State police agencies were obtained from the 2000 Census of State and Local Law Enforcement Agencies. Additional state-level investigative agencies were derived from previous NDTS sampling plans. Typically included for each state were the state police and lead drug enforcement agency, although this pattern varied in some states (stratum 98).
   
• Investigative agencies in three U.S. territories--Guam, the Northern Mariana Islands, and Puerto Rico--were also selected with certainty (stratum 99).

To ensure that state-level representative statements could be made about results obtained from the NDTS 2003, local law enforcement agencies were coded according to the 50 states and District of Columbia. Municipal police departments and county sheriff's offices with sworn FTE employees of 10 or more but fewer than 75, and meeting all the criteria discussed above, were included in these strata.²² The states were used as the noncertainty strata, and a Neyman allocation was used to allocate the noncertainty sample to the state strata. All eligible law enforcement agencies in the District of Columbia and Hawaii met the criteria for inclusion with certainty and were included in stratum 97. The state of California was split: law enforcement agencies within the Southern and Central U.S. Attorney Districts were included in Southern California and those in the Eastern and Northern Districts were included in Northern California. The noncertainty agencies in Southern California were included in stratum 91, and similar agencies for Northern California were included in stratum 92.

The actual sample, representing the sampling universe of 7,930 state and local law enforcement agencies, consisted of 3,497 agencies in 53 strata, 3 of which were certainty strata.

To Top     To Contents

Data Collection

NDIC verified the point of contact and mailing address for each law enforcement agency in the sample and mailed the surveys, which were accompanied by a cover letter from NDIC Director Michael T. Horn and a postage-paid return envelope. NDIC Field Program Specialists located throughout the country were responsible for follow-up contacts with sample agencies that were mailed a survey.

Of the 3,497 state and local law enforcement agencies in the actual sample, 251 had received the survey earlier in 2003 under a joint effort by NDIC and the High Intensity Drug Trafficking Area (HIDTA) program that was designed to assist the HIDTAs in preparing their annual threat assessments. Copies of surveys completed by sample agencies under the joint NDIC-HIDTA effort were forwarded to NDIC. Lists of agencies that did not respond were given to Field Program Specialists for follow-up contact, and a second NDTS 2003 survey was either mailed or personally delivered to the nonresponding agency.

NDIC provided daily reports to help Field Program Specialists target nonresponding agencies, which were contacted by telephone, by letter, and in person. All responses were entered in the NDTS database designed and developed by NDIC.

To Top     To Contents

Sample Adjustments

During survey processing, NDIC identified nine ineligible agencies. Included in this group were five agencies that no longer performed drug enforcement activities, three agencies that no longer existed, and one agency that had merged with another law enforcement agency. Three of these agencies were state noncertainty cases (one each in stratum 24, stratum 29, and stratum 91), three were certainties due to size (stratum 97), and three were state agency certainties (stratum 98).

The nine ineligible agencies were deleted from the original actual sample of 3,497 that resulted in an adjusted sample of 3,488 agencies in 53 strata, three of which were certainty strata. The sample represents 7,921 agencies. A summary of the adjusted sample design is presented in Table A1.

To compensate for the deletion of the three ineligible records in noncertainty strata (stratum 24, stratum 29, and stratum 91) from the sample, a poststratification factor was calculated for the affected strata to correct the base weights for those strata. For all other strata, the poststratification factor is 1.0. The poststratification factors for all strata also are shown in Table A1.

To Top     To Contents

Nonresponse Adjustment Factor

Of the 3,488 agencies in the adjusted sample, 3,354 agencies responded to the NDTS 2003 for an overall response rate of 96.2 percent. Table A2 on page 107 summarizes the response rates by state. A nonresponse adjustment factor was applied to account for those agencies that did not respond to the survey.

The nonresponse adjustment factor for each stratum j is calculated as

d-link

where k represents either the k^th responding or the k^th nonresponding agency in stratum j.

The final weight for each responding agency is calculated as

d-link

To Top     To Contents

Estimation Techniques

The final weight for each respondent was used to derive national, regional, and state-level estimates for all survey items. The final adjusted score was summed for each response category (for example, high, moderate, and low) for each item, and the proportion of the final scores provided the national, regional, or state-level estimate for that item. Some respondents did not answer all survey items. The item nonresponse rate ranged from 0.8 to 20.3 percent.

To Top     To Contents

Nonsampling Error

Nonsampling error may affect NDTS 2003 data. Possible nonsampling errors include the following:

• Inability to obtain information about all agencies in the sample
   
• Varied interpretation of response categories (for example, high, moderate, and low are defined differently by respondents)
   
• Inability or unwillingness of respondents to provide correct information
   
• Errors made in collection, coding, or processing of data
   
• Failure to represent all agencies within the sample (undercoverage)

Nonsampling error can increase the total error over the error resulting from sampling. Random nonsampling errors can increase the variability of data, while systemic nonsampling errors that are consistent in one direction can introduce bias into the results of a sample survey. NDIC used data collection, coding, and processing procedures designed to limit the effects of random nonsampling error on the NDTS 2003 data. No systemic nonsampling errors were identified.

To Top     To Contents