Supporting Statement_section_b

Department of Commerce

U.S. Census Bureau

OMB Information Collection Request

State and Local Government

Finance Forms

OMB Control No. 0607-0585

Part B. Collection of Information Employing Statistical Methods

1. Universe and Respondent Selection

The 2017 Census of Governments: Finance will canvass the entire universe of governments of about 77,000. For our Annual Survey of State and Local Government Finances, we use an estimated sample size of about 11,000 state and non-school local governments selected with probability proportional to size in order to estimate data for governments other than school districts¹. For the 2018 and 2019 government finance annual surveys, the estimated sample distribution for non-school local governments will be as given in Table 5 supplemented with samples of births each year.

Table 5. Sample and Universe Counts of Non-Education Governments
Government Type	2017 Census of Governments (Universe)	Estimated Sample (For Fiscal Year 2018 Survey)
States	50	50
Counties	3,031	1,781
Municipalities	19,519	3,507
Townships	16,360	1,066
Special Districts	38,266	4,503
Total	77,226	10,907

The estimated sample of 10,857 non-school governments for the 2018 annual survey is a state-by-type government sample that will be selected from the universe of the 2012 Census of Governments. The sample of non-school governments for the 2019 annual survey is also a state-by-type government sample that will be selected from the 2017 Census of Governments. Samples for the annual surveys are selected every five years with new samples beginning in years ending in ‘4’ and ‘9’.

The unit response rates for the 2014 Annual Survey of Local Government Finances, our most recent completed survey, were:

• for county governments 77.1%

• for cities 79.0%

• for township governments 74.8%

• for special district governments 75.9%

The Annual Survey of State Government Finances is a full data canvass of all 50 states and dependent agencies. The unit response rate for 2014 was 100%.

For the data collection concerned with collecting information from public employee pension systems, Forms F-11 and F-12, we will use a sample of 2,075 state and local pension systems selected with probability proportional to size in order to produce state area estimates at the state and local level. This sample is supplemented each year with a sample of births.

Table 6. Sample and Universe Counts of Public Employee-Pension Systems
System Type	Universe for 2017 Census of Governments	Estimated Sample for 2018 & 2019 Annual Surveys
State Systems	299	299
Local Systems	6,475	1,774
Total	6,774	2,073

The overall response rate for the 2015 Annual Survey of Public Pensions was 65.3%. The weighted response rate was 98.6%.

2. Procedures for Collecting Information

Sampling

The Annual Survey of Local Government Finances for the 2018 - 2019 surveys will use a state-by-type of government stratified probability proportional to size sample drawn with Census of Governments data. The 2018 sample will be the last one based on the 2012 Census of Governments, and the 2019 sample will be a redesign based on the 2017 Census of Governments. The redesigned sample size will be determined by the data quality required by the quality standards and the funding allocated by the survey.

The initial certainty criteria for the 2018 sample are as follows:

1) All local governments in the District of Columbia and Hawaii;

2) All county governments with a 2012 population of 500,000 or more;

3) All cities with a 2012 population of 200,000 or more.

Note that the initial certainty criteria for the 2019 sample redesign could be revised depending on changes due to future population growth.

Non-certainty units are selected as follows:

For the Annual Survey of Local Government Finance a size variable is defined for the sample selection. The size variable for the finance survey is the maximum of total expenditures and a ratio-adjusted second variable, which depends on the local government type. For county governments, the total taxes is used; for cities and townships, the total revenues is used; and for special districts, the total long-term debt is used. Excluding all certainties, we group remaining government units by the unit size variable. If the size variable is zero, then they belong to the “no activity” stratum. Otherwise, they belong to the stratified probability proportional (πPS) universe. A πPS sample is conducted by state and government types. For some townships and special districts, we further apply a modified cut-off sample methodology, where subsampling is performed to reduce the number of non-contributory sub-counties and special districts in the sample.

 

Between sample redesigns, we supplement the sample with births and remove disincorporation, or deaths from the sample. We take all general purpose governments and add them as births to the sample. In years with a large numbers of special district births, we apply a systematic sample method by function code and state and then add the selected births to the sample; otherwise, all special district births are added to the sample.

 

The Annual Survey of Public Pensions for 2018 - 2019 will use a state area by level (state-local) sample drawn from Census of Governments (CoG) data. The 2018 sample will be the last one based on the 2012 CoG, and the 2019 sample will be a redesign based on the 2017 CoG. The initial certainty criteria for the 2018 sample are as follows:

1) All state government pension systems;

2) Local systems with total holdings and investments of $500 million or more;

3) All Local pension systems in states with 20 or fewer state and local pension systems.

Note that the initial certainty criteria for the 2019 sample redesign could be revised depending on changes due to future population growth.

The total holdings and investments is defined as the size variable whenever it is available. Local governments with this variable are selected for the sample based on a probability proportional-to-size design without replacement (πPS) for each state and government type. For units that are missing total investment and holdings, the ratio of a proxy variable to the total employees is used to construct strata using the cumulative square root of the frequency method.

Estimation Procedure

 

The survey calculations will employ a model adaptive method to estimate aggregated level such as general sales, selective sales, liquor stores revenue, current operation, capital outlay, education, health, highways, and so on, and then apply an indirect synthetic estimation method to project the total for detailed viewable and downloadable variables. Pension Survey estimates are developed using a simple design-based unbiased estimator, the Horvitz-Thompson.

Degree of Accuracy

 

State Area Estimates of Local Government Totals. For state area estimates of only local government totals, the finance sample is designed to produce totals with a relative standard error of three percent or less or five percent or less on the following 11 major financial items:

Three percent or less:

• Long-term debt outstanding

 

• Total revenue

 

• Total expenditure

 

• Criminal justice expenditure

Five percent or less:

 

• Education expenditure

 

• Highways expenditure

 

• Health expenditure

 

• Housing expenditure

 

• Total capital outlay expenditure

 

• Utilities expenditure

 

• Welfare expenditure

State Area Estimates of State and Local Government Totals. With the addition of state government data to the local government information, the finance sample for survey generally yields state area estimates with a relative standard error of one percent or less on total revenue and total expenditures and under two percent for long-term debt outstanding.

 

State Area Estimates of State and Local Pension Systems. For state area estimates of state and local pension systems, the sample is designed to produce totals with a relative standard error of three percent or less on total cash and security holdings of public employee pension systems.

Unusual Problems

None.

Use of Periodic Data Collection Cycles

These data are critical components for calculating the National Income and Product Accounts produced by the Bureau of Economic Analysis, which require a minimum of annual data.

3. Methods to Maximize Response

We use several methods and resources to maximize response in the collection of government finance data, including: follow-up mailings; telephoning nonrespondents; central data collection agreements; and Internet data collection.

4. Tests of Procedures or Methods

Over the past several years, the Census Bureau has conducted record keeping practice studies and cognitive testing to examine the financial activity of state and local governments, and public employee pension systems. The results of the cognitive testing and record keeping practice studies have been incorporated into survey questionnaires where feasible. The Census Bureau will continue to test survey questionnaires as needed when planning questionnaire changes.

5. Contacts for Statistical Aspects and Data Collection

Bac Tran (301-763-1937) Chief, Public Sector Sample Design and Estimation Branch, monitored the statistical aspects of the sample design. Randy Moore (301-763-7231), Chief, Local Finance Statistics Branch, International Trade Management Division, is available to answer any data collection questions.

List of Attachments

1. Initial Request and Follow-Up Letters

2. Supplemental data letters

(3) Forms F-5 Spreadsheet, F-11, F-12, F-13, F-28, F-29, F-32

(4) Federal Register Notice response letter

(5) Electronic reporting instrument login screen and burden statement for local finance (F-28,
F-29, F-32) and public pensions (F-11 and F12).

1¹ Elementary and secondary school system finance data are collected under authority of a separate submission; see OMB clearance 0607-0700, expiration date 10/31/2018 through a joint undertaking between the National Center for Education Statistics and the Census Bureau. For the 2017 Census of Governments, we will use the universe data and for the annual surveys in 2018 - 2019, we will use the universe for independent school systems and only the dependent school systems included in those government units chosen in the non-school local government sample.

7