Data Dictionary
Summary:
1. Title and Project Information
2. Background
3. Overall Question
4. Origin of Data
5. Data Parameters
*************************************
Project: Analysis of County-Level Discrimination in Federal Marketplace
Authors: Doug Jacobs, Nicolas Barcelo, Alicia Fernandez, Isabel Allen
Institution: UCSF School of Medicine
*************************************
Background:  According to federal regulations, health insurance service areas must be established without regard to racial, ethnic, language, health status-related factors, or other factors that exclude specific high utilizing, high cost, or medically underserved populations.  

	Recent research has suggested that some insurers may not be adhering to these requirements. A 2015 Commonwealth Fund report on insurance competition in the first enrollment period generated maps that suggested that, in some states, the number of insurers selling their plans in a county depends on relative population size and per-capita income of the rating area.  This is potentially problematic, because if insurers are using demographic variables to decide whether or not to sell insurance in a county, certain underserved populations may be exposed to fewer insurance choices. This, in turn, may lead to higher premiums for these populations, because less competition among insurers raises premiums. 
	 
	To date, no research has clarified what factors correlate with insurers selling their plans in a county. As such, we undertook a comprehensive analysis of all counties in the federal marketplace, comparing Census Data with the number of insurers selling their plans in each county. Our specific focus was on historically underserved population, ie racial/ethnic minorities, the previously uninsured, and low income populations.

*************************************
Question: 
1. Do county-level demographics predict the cost and availability of insurance plans in the federal marketplace?
2. What are the important covariates affecting cost of insurance.
3. Do pre-Obamacare states (MA & CA) differ significantly from the rest of the country.
4. Are states with a majority Black or Latino population different from other states.

*************************************

Data:
Composed of:

1. Health Insurance Data (2015) by State and County for ALL states participating in the federal marketplace
	- Number of Insurance companies entering county-based market
	- Average Silver-level premium cost / county
2. Census Data
	- demographic variables corresponding to each county from American Community Survey Summary (2010-2014)
	
*************************************
	
Data Parameters:

	2511 rows = counties
	
*************************************

Defining Variables: 

“ID” - ID number of county

"State" - abbreviated  
    
"COUNTY" - full county name  
      
"Co_State" - county and state name  
   
"Issuers" - number of issuers / county      

"SilverPlans" - number of silver plans / county

"PremiumMean" - mean cost of silver plan premium / county

"TotalCPop" - total population / county in thousands

"WhiteP" - % white / county       

"BlackP" - % black / county       

"LatinoP" - % hispanic / county

“Total population age groups” - Under5 - Age 60-64 in thousands

"LHSP" - % of persons with less than high school education / county

"MHI" - Median Household Income / county

“MHI1” - Median Household Income / county divided by 1000
          
"EmpP" - % of employed persons / county

"UnInsP" - % of persons uninsured / county

*************** the following columns have been generated and are admittedly experimental 
 
voluntary is a binary variable to categorize states mandating insurer participation in all counties (=0) vs "voluntary" participation in counties (=1) 

mandatory is a binary variable applicable to states (=0) where ALL counties will have same # of issuers vs. in states (=1), counties will have a range of issuers.

latino state is a binary variable indicating a latino state.

*************************************