Supporting Statement Part B

SUPPORTING STATEMENT

Part B: Collection of Information Employing Statistical Methods

1. Respondent Universe and Sampling methods:

CCI will create a representative sample of law enforcement agencies based on data available from the National Directory of Law Enforcement Agencies (NDLEA) from the National Public Safety Information Bureau. From the NLDEA directory, the full list of leaders from the Municipal Law Enforcement and County Law Enforcement will be obtained. The combined lists report approximately 15,549 law enforcement agencies across the country, including municipal police and county law enforcement agencies. In order to achieve the twin goals of a drawing a nationally representative sample of law enforcement agencies and forming the basis of a “working” compendium of police-led diversion programs, we will employ a state-based stratified sampling technique. Our overall sample size will be 25% of the total. Most serve small jurisdictions (i.e. city or county populations of fewer than 25,000 residents), while just 74 agencies (less than 1%) serve populations greater than 250,000. Although there is no evidence that agencies in larger jurisdictions or those in particular states or regions are more likely to develop diversion programs, because of the unique nature and small number of large urban jurisdictions, we will include all of these agencies in the sample. Thus, CCI will utilize a three-step sampling technique. First, we will calculate the percentage of total law enforcement agencies represented by each state (e.g., 440 law enforcement agencies in New York State= 3% of the total 15,000 agencies). Next, we will draw a sub-sample equal to this percentage for each state to form our full sample of 3,600 (e.g., our subsample for New York State will be 108). Finally, we will sample the necessary number of agencies for each state based on two pre-defined strata: a complete sample of all agencies in the state serving a jurisdiction of 250,000 or more and a random sample of all other agencies (serving a jurisdiction of less than 250,000) that includes a sufficient number of agencies to reach the state’s total target number. We will assume a 75% response rate in jurisdictions of less than 250,000 and a 90-100% response from the 74 large agencies (they will receive particularly intensive follow-up).

2. Procedures for the Collection of Information

CCI will subcontract with a professional survey research firm to administer the survey. We will conduct a competitive bid process to obtain the most experienced and efficient research firm to administer the survey that CCI develops. CCI will collaborate with the survey firm to develop the most efficient and comprehensive administration strategy and timeline. The initial invitation to participate in the survey will request that the law enforcement agency assign the individual best positioned to answer questions about police-led diversion. Each agency will only be allowed to complete one survey and the respondent will be able to change responses at any time until the survey closes.

3. Methods to Maximize Response Rates and Deal with Issues of Non-Response

Administration will include follow-up until a response rate of close to 75% is reached for smaller agencies and close to 100% for the 74 largest agencies. Additionally, approximately halfway through the four-month administration phase, a second set of surveys will be administered to any agencies not collected in the original sample that were reported by original respondents to be running a police-led diversion program. Survey responses from agencies that are added to our sample in this fashion will not be used for statistical purposes (so as to avoid biasing our initial representative sample), but we will add these responses to our working national compendium. In this way, we expect that our survey process will identify well in excess of 25% of existing police-led programs nationwide.

After initial data cleaning, survey completers will be compared to non-completers, based on agency characteristics (e.g., size, region of the country and state) and adjusted for non-response bias. Prior to analysis, the final survey sample will be weighted such that it is representative of the national population of all agencies. Weights will be calculated as the inverse of the probability of selection. The sampling weights will then be standardized (mean weight = 1) and adjusted to compensate for any differences in non-response.

4. Tests of procedures or methods to be undertaken

Analysis of frequency distributions will be run to describe police-led diversion program goals, policies, and operations nationwide. Relationships will be examined between agency size and region/state of agency with presence of a police-led diversion, program subtype (i.e., CIT program, juvenile, or adult program), and other program/policy characteristics; relationships between police-led diversion program goals and specific policies and practices will also be examined. The confidence interval will be +/- 1.7% for the entire sample (assuming n=2700, alpha=.05, and prevalence=.50). For subgroup analyses of the 74 large jurisdictions, assuming that we obtain a survey from 68 (just over 90%), the confidence interval will be +/- 3.7%. For parameters whose prevalence is farther from .50 there will be even narrower margins of error.

5. Individuals consulted on statistical aspects of the design and organization/persons collecting and analyzing the data.

No individuals outside of CCI or any other organizations will be consulted on statistical aspects of the design and analyzing the data.

2