• No results found

Quantitative Data Sources

In document 5554.pdf (Page 53-58)

CHAPTER 3. Research Design and Methods

3.2 Quantitative Data Sources

State Medicaid programs use two forms to report expenditure data. States use Form 64 to claim reimbursement for Medicaid care. Alternatively, states use Form 372 to declare cost neutrality for HCBS. For the spending per capita analysis for aims 1 and 2, I use the Form 64 data, since more years are available and other state-level HCBS studies use these data. These data are collected by Eiken and colleagues (2011) and housed on www.hcbs.org.4 Please refer to table 3.1 for a full list of the quantitative measures.

Form 372 data providing HCBS recipient and expenditure data are publicly available from the University of California San Francisco Center for Personal Assistance Services (UCSF PAS). Researchers at UCSF PAS collect the data yearly via telephone interviews with state Medicaid officials. The data are housed on the Kaiser Family Foundation website. UCSF researchers collect expenditure and recipient data for all three types (home health, PCS, and waiver) of Medicaid long-term care HCBS programs (Howard, et al., 2011). I use form 372 data for the recipients per capita and spending per recipient analysis in aim 1. Form 372 is the only source of recipient data. I use the

expenditure data provided by this form, not data from Form 64, to create the expenditures per recipient variable as it would be unwise to use data from two different sources to

                                                                                                               

4   I am incredibly thankful to Steve Kaye for sharing his reformatted Form 64 data with me.  

create such a variable. I have permission to use these data for years 1999-2008. I adjusted all expenditure these data for inflation using the Consumer Price Index (CPI) available from the Bureau of Labor Statistics.5

Centers for Medicare and Medicaid Services

Data for home health users are from the Centers for Medicare and Medicaid services (CMS). The home health user data were divided by each state’s population to obtain the users per 1000 state population proportion (Centers of Medicare and Medicaid Services (CMS) "Medicare and Medicaid Statistical Supplement, 2012). Data for the Federal Medical Assistance Percentages (FMAP) are from CMS through the CMS website (ASPE, 2012).

State Political Ideology Data

Political scientist William Berry collects and makes available political ideology data (Berry, Ringquist, Fording, and Hanson, 1998). The citizen ideology variable is computed by adding the product of the Congressional incumbent’s ideology and the vote share he or she received in the last election to the product of his or her challenger’s ideology and vote share in the last election.6 The ideology of the incumbent is measured by the Americans for Democratic Action (AFDA). The ideology of the challenger is assumed to be equal to the average ideology score of all the incumbents in the state for

                                                                                                               

5 I use the regional medical care services CPI calculation found at: http://www.bls.gov/cpi/

6 Citizen Ideology=(Incumbent Ideology)(Incumbent Vote Share)+(Challenger Ideology)(Challenger Vote Share)

the same party. This measure is then complied for a state-wide measure. The ideology ranges from 0-100 with 100 being the most liberal. These data are publicly available from Professor Richard Fording’s website (Fording, 2012).

Book of the States

The Book of the States, a yearly state almanac, provide state-level political data for party control of state governorships and state legislatures (Book of the States, 1995, 1997, 1999, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008). These measures are indicator variables that identify which party controls the governorship and legislature for each year.

United States Census Bureau

The Statistical Abstract, a yearly publication of state-level characteristics, provide data on economic and socio-demographic data (US Census Bureau The 2012 Statistical Abstract: Earlier Editions, 2012). The population of each state over 85 years old, each state’s African American population, and each state’s Hispanic population are divided by the state’s respective total population for each year and multiplied by 100 to create a percentage. Each state’s per capita income is adjusted for CPI, as are the expenditure variables. Each state’s female labor force participation rate is available from the yearly statistical abstract as a percentage. The poverty rate variable is available through the Small Area Income and Poverty Estimates for each state. These measures are yearly percentages (Small Area Income and Poverty Estimates, 2012). The measure for each state’s metropolitan population was calculated from the measure from the Office of

Management and Budget and the Census Bureau. The state percentages were calculated using the following procedure: the population of the metropolitan counties were tabulated and then divided over the total population for each state (Metropolitan and Micropolitan, 2012). The data for state population over age 65 with a self-care limitation are from the American Community Survey kept by the Minnesota Population Center (Minnesota Population Center-Home of the IPUMS and other Data Projects, 2012).

Online Survey Certification And Reporting (OSCAR)

OSCAR data were used to obtain nursing home bed censes. These data are reported in the serial publication Health, United States (2012). The numbers of beds per state were divided by the state population to create the nursing home beds per 1,000 state population statistic.

Area Resource File (ARF)

The ARF provides home health agency information (Health Resouces and

Services Administration, 2010). The number of home health agencies was divided by the state’s population to provide the agencies per 100,000 state population variable.

Medicaid Nursing Home Reimbursement Rate Data

Nursing home reimbursement data were obtained from the authors of previous work (Grabowski & Gruber, 2007) and were adjusted using medical care CPI. These data are only available for years prior to 2004.

Bureau of Labor Statistics

The United States Bureau of Labor Statistics (BLS) holds data for average wages for states for the past several years. I use the Occupational Employment Statistics Query System to include a home health aide wage variable (code 31101) in the quantitative analysis (Occupational Employment Statistics Query System, 2012). The state

unionization variable is a percentage obtained from the Bureau of Labor Statistics (Union Affiliation of Employed Wage and Salary Workers by State, Annual Averages 2011).

Center For American Women And Politics

The Center for American Women and Politics holds data for women in state legislatures for the last several decades (Center for American Women and Politics, Archived Fact Sheets, 2012). I use their data to create a variable denoting percentage of legislators that are women.

Missing Data

Data were missing for some years for select independent variables and were assumed to be missing at random. I used linear interpolation to address missing data. Thus, for each state, I regressed any variable with missing data on years and imputed the missing data as appropriate. Appendix B provides more detail concerning what variables are missing.

3.3 Sample Size and Statistical Power for Quantitative Analysis

In document 5554.pdf (Page 53-58)

Related documents