The Census Bureau released the research Supplemental Poverty Measure (SPM) estimates in November 2011 using data from the Annual Social and Economic Supplement of the Current Population Survey (CPS ASEC).  However, since the Bureau recommends the use of the American Community Survey (ACS) for poverty estimates for sub-national geographic units, it is important to explore how this new measure can be estimated from ACS data. The challenge is that the ACS is missing a number of key data elements required to produce SPM estimates, including program participation, the value of SNAP benefits, taxes, child care expenses, medical out of pocket expenditures and some relationship codes. This paper will explore how these data limitations might be overcome. 

On April 1, 2011, the Census Bureau sponsored a workshop at the Urban Institute on State Poverty Measurement Using the American Community Survey. (For a summary of the workshop see http://www.urban.org/publications/412396.htm.l)  This workshop was designed to provide expert advice to the Census Bureau on the development of a

SPM at the state level using the ACS. The workshop participants discussed the myriad challenges involved in using the ACS to produce SPM estimates.  The summary of the workshop concludes with the following text:

The conversation confirmed that producing state-level estimates of the SPM using the ACS is a challenging undertaking. There are no easy shortcuts. Nonetheless, researchers agreed that it is not in our collective interests for multiple variants of the SPM to spring up. Most agreed that ideally one institution would take charge of creating a common framework for implementing the SPM on the ACS. Workshop participants hoped that the Census Bureau would play this role. No other institution has the range of expertise, resources, and access to confidential data to undertake this task.

This paper is an attempt to be responsive to that conversation.  Building on the extraordinary work at the state and local level using the ACS to develop SPM-like estimates, this paper will continue to develop a proposal for how the Census Bureau might produce SPM estimates using ACS data.  The paper will have a particular emphasis on how the data in the CPS ASEC might be used to inform the ACS imputations.  While preliminary work on this topic relied exclusively on econometric modeling, this paper will explore the feasibility of a statistical match between the two surveys.  In addition, this paper will discuss the pros and cons of including some of the more difficult expenditure items in the thresholds.  A second focus of this exploration is to assess the feasibility of producing an ACS public use research file that researchers could use to produce local SPM estimates that are comparably calculated for all areas.  Due to this second focus, the analysis in this paper will use the 2010 ACS PUMS file.