Section B

Supporting Statement

U.S. Department of Commerce

U.S. Census Bureau

Quarterly Survey of Public Pensions

OMB Control Number 0607-0143

(F10)

Section B. Collections of Information Employing Statistical Methods

1. Description of Universe

The 100 largest public retirement systems, as determined by their total cash and security holdings reported in the 2007 Census of Governments, account for about 90 percent of the total assets of all state and local government retirement systems. These 100 systems are the universe for the survey.

After a census has been taken, it is considered best practice to reselect the largest 100 retirement systems. The first quarter for 2009 presented data from the new universe of the 100 largest public retirement systems. The previous quarters represent data from the old universe. Five of the 100 units from the old universe were replaced with new retirement systems. The Census Bureau released a report online that contains the bridge between the old and new units for the fourth quarter of 2008 and the first quarter of 2009. The universe will be reexamined after the 2012 Census of Governments. Any changes to the universe will be implemented for the first quarter of 2014 with another bridge report to provide data users with a link between the old universe and the new universe.

This survey does not produce any statistical estimates for the public-employee retirement systems universe.

2. Sampling Methodology and Estimating Procedures

The data in this survey are from all 100 units in the target population. Therefore, all weights are 1.0000, and the estimates are simple aggregations of reported and imputed data. On average, 12 percent of the data are imputed.

Starting with the data release for the first quarter of 2011, a new methodology was used for imputing data for retirement systems that did not respond to the survey. A bridge study was published to provide data users with a link between data based on the old imputation method and data based on the new method.

3. Methods to Maximize Response

Of those who respond to the survey, 89 percent are from the Web. The Census Bureau is planning to only mail survey forms to the six respondents who continue to fax and mail in their responses and to the five units who continue to be non-respondents. All of the 100 respondents will receive an email reminder encouraging them to report online.

In an effort to maximize response Census Bureau staffs have initiated the following over the past three years:

1. Earlier mail-out:

As of 2010, the Census Bureau mails the survey forms at the end of the calendar quarter. Prior to 2010, the survey forms mailed four weeks after the end of the calendar quarter. See Attachment 3.

2. Improved telephone follow-up:

Staffs use prior response rate analysis to determine which units report when, focusing follow-up efforts on the early reporting respondents in the early weeks and later reporting respondents in the later weeks, thereby attaining response rate goals methodically and more efficiently.

3. Efficient data compilation:

Staffs research which public pensions publish financial data on the Internet and maintain a record of these systems. If these systems do not report by data collection closeout, staffs compile data from the Internet for these particular units. Late reporting units use the option to submit late data for prior quarters as well as submit revisions to prior quarter data. Revisions are made to the prior quarter data releases to incorporate these late receipts and revisions.

4. Established new contacts:

Staff attended conferences and met with representatives from public retirement systems to establish new contacts for previous chronic non-reporting systems. At the 2010 National Association of State Retirement Administrators (NASRA) Annual Conference, Census Bureau staff met with representatives of several state systems in the Quarterly Survey of Public Pensions panel to highlight the importance of these data and to encourage their participation. New contacts were established with state administered retirement systems in Hawaii, South Carolina, and Wisconsin – three of the six previously chronic non‑respondents. These systems began reporting in 2011.

The response rate has improved since 2007 (when it reached 82 percent). Since the beginning of 2009, the response rate has remained between 85 and 95 percent. Most notably, however, is the response rate attained in the earlier weeks of the survey cycle. In the second quarter of 2009, the response rate during week 13 of the survey cycle was only 8 percent. In the second quarter of 2011, the data were released during week 13 with a response rate of 87 percent.

Prior to the first quarter of 2010 release, data were released 17 weeks after the reference period. As of the third quarter of 2010 release, the data are released 13 weeks after the reference period without compromising the response rate. See Attachment 3. It is the higher response rate in the earlier weeks that has enabled the Census Bureau to release four weeks earlier and to provide preliminary data to both the Bureau of Economic Analysis (BEA) and the Federal Reserve Board for GDP projections and the Flow of Funds Accounts respectively. Prior to 2010, the response rate was too low to provide preliminary data to BEA and the Federal Reserve Board.

3. Testing of Procedures

The frequency, limited scope, flexibility, and continuity of this survey make it essentially self-testing. The Census Bureau has conducted this survey since 1968 and the content has changed very little over the past 20 years. During both technical and non-response follow-up calls, we emphasize our flexibility concerning formats and procedures and encourage the survey panel to report electronically to facilitate their responses.

We recently conducted cognitive testing in conjunction with the Response Improvement Research Staff (formerly the Establishment Survey Methods Staff) at the Census Bureau. Multiple respondents were interviewed regarding their understanding of the questionnaire and their ability to provide data on the current survey form, as well as the new and modified questions. It was through this cognitive testing that the Census Bureau determined it was feasible to break out the “Net gain or loss on investments” question into three parts for added granularity.

5. Contact

Contact the person listed below for statistical aspects and data collection:

Erika H. Becker-Medina

Chief, Employment & Benefit Statistics Branch

Governments Division

(301)763-1494

erika.h.becker.medina@census.gov

Carma Ray Hogue

Assistant Division Chief, Statistical Research and Methodology

Governments Division

(301)763-4882

carma.ray.hogue@census.gov

Attachments

1. Form F-10

2. Initial Mail Letter

3. Timeliness and Response Rates Graph

3