Millions of Hospital Visits Tell us About the Value of Public Health Insurance as a Safety Net for the Young and Privately Insured Amanda E Kowalski Yale University and NBER February 2015 ID: 935356
Download Presentation The PPT/PDF document "What D o Longitudinal Data on" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
What Do Longitudinal Data onMillions of Hospital Visits Tell us About the Value of Public Health Insurance as a Safety Net for the Young and Privately Insured?
Amanda E. Kowalski
Yale University and NBER
February 2015
Slide2What Do Longitudinal Data on Millions of Hospital VisitsTell us About the Value of Public Health Insurance as a Safety Netfor the Young and Privately Insured?Young, privately insured individuals with hospital visits for diagnoses that require more hospital visits in future years are more likely to transition to public health insurance in future yearsIf we ignore the longitudinal transitions in the data, we miss over 80% of the value of public health insurance as a safety net
Slide3What Do Longitudinal Data on Millions of Hospital VisitsTell us About the Value of Public Health Insurance as a Safety Netfor the Young and Privately Insured?What the prior literature does not tell usWhat my data can tell us – stylized factsHow I incorporate my data into a framework for valuing public health insurance
What we learn from the framework, along with robustness
The longitudinal value of health insurance is much greater than the sum of its cross-sectional parts
Slide4The current private health insurance system does not offer long-term protection against financial riskI consider the value of public health insurance as a “safety net” that fills gaps in the current private systemPrevious literature focuses on design and regulation of private health insurance contracts to mitigate this riskCochrane (1995), Pauly
(1995)
I use longitudinal data and a longitudinal framework
Previous literature focuses on the value of public heath insurance for those who have already been caught by it, using cross-sectional data and a cross-sectional framework
Medicare: Finkelstein and McKnight (2008),
Khwaja
(2010),
Barcellos
and Jacobson (2014)
Medicare Part D:
Engelhardt
and Gruber (2011)
Medicaid: Finkelstein,
Hendren
,
Luttmer
(2014)
Slide5What Do Longitudinal Data on Millions of Hospital VisitsTell us About the Value of Public Health Insurance as a Safety Netfor the Young and Privately Insured?What the prior literature does not tell usWhat my data can tell us – stylized factsHow I incorporate my data into a framework for valuing public health insurance
What we learn from the framework, along with robustness
The longitudinal value of health insurance is much greater than the sum of its cross-sectional parts
Slide6I use longitudinal data on almost all hospital visits in New York from 1995-2011Using my SPARCS data and population data, I create a balanced panel to represent New York State populationI focus on individuals who are young and privately insured in 1995 to isolate the value of the safety net
Slide7Just for the stylized facts, I focus on individuals with “persistent diagnoses,” likely to drive the value of private health insurance as a safety net
Slide8Young, privately insured individuals with persistent diagnoses have higher costs in future years, and they are more likely to transition to public insurance in future yearsEven after imposing an annual upper bound of $30K, cumulative costs are $58K by 2011 for persistent diagnoses vs. 13K for other diagnoses
17.9% of individuals with persistent diagnoses have public coverage in 2011, in contrast to 3.7% with other diagnoses
Slide9What Do Longitudinal Data on Millions of Hospital VisitsTell us About the Value of Public Health Insurance as a Safety Netfor the Young and Privately Insured?What the prior literature does not tell usWhat my data can tell us – stylized factsHow I incorporate my data into a framework for valuing public health insurance
What we learn from the framework, along with robustness
The longitudinal value of health insurance is much greater than the sum of its cross-sectional parts
Slide10Simple indifference condition for valuing public health insuranceRooted in frameworks used in the literatureWith public insurance Without public insurance
Closed form solution for
ρ
under CARA
Slide11Goal: calculate how much we miss by using cross-sectional in lieu of longitudinal dataSpecial Cases of the Value of InsuranceRatio of interest
Slide12Contrast to other frameworks that produce the same value with cross-sectional or longitudinal dataKowalski (2015)Handel, Hendel, and Whinston (2013)
The cross-sectional value of insurance from our framework
φ
is almost the same as the alternative cross-sectional value of insurance
μ
Slide13Implement the indifference condition empiricallyCosts in actual world with public insurance: where:Costs in counterfactual world without public insurance:
where:
Uses all paths observed in the data, people with public coverage either gain private coverage or go uninsured
Examine robustness to wide range of parameters
Slide14What Do Longitudinal Data on Millions of Hospital VisitsTell us About the Value of Public Health Insurance as a Safety Netfor the Young and Privately Insured?What the prior literature does not tell usWhat my data can tell us – stylized factsHow I incorporate my data into a framework for valuing public health insurance
What we learn from the framework, along with robustness
The longitudinal value of health insurance is much greater than the sum of its cross-sectional parts
Slide15Longitudinal value of insurance much larger than cross-sectional value for a large range of risk aversion parametersRed dashed line shows traditional framework that gives the same answer for longitudinalAnd cross-sectional data
Slide16Robustness to Parameters: αAcross a broad range, lose over 90% of value
Even though the levels vary a lot with
α
, the ratio stabilizes
Slide17Rather than calibrating one value, consider robustness to a large rangeAll necessary parametersα, coefficient of absolute risk aversionM, upper bound for annual individual costsT, number of years of data in sampleN
, percentage of population in sample
Private Information on Persistent Diagnoses
Robustness to Including Uninsured
Slide18Robustness to Parameters: MAcross broad range, lose over 90% of value
At very small values of M, there is not much risk, so all three values are close.
We choose 30K as our baseline value because that is where the ratio stabilizes.
As M gets really large, variability within a period increases while mean changes little,
so cross-sectional value increases relative to risk-neutral value
B
oth are small relative to longitudinal value – frequent visits dominate expensive visits
Slide19Robustness to Parameters: TAcross broad range, lose over 90% of value
Use all years as our baseline value.
Ratio stabilizes after about 8 years. MEPS only has 2.5 – not long enough for values to diverge.
Slide20Robustness to Parameters: NAcross broad range, lose over 90% of value
Full sample contains 1.69 million individuals.
MEPS is approximately 0.3% of our sample. Results at that size are highly variable.
Results from 100 draws of the size of the MEPS range from 8.2% of 95.7% - too small for tails!
Furthermore, this calculation assumes MEPS has 17 years of data, but it only has 2.5!
Slide21Robustness to Private Information on Persistent DiagnosesEven with extreme private info, results persist
Slide22Robustness to Including UninsuredThese are baseline, somewhat less intuitive
Slide23What Do Longitudinal Data on Millions of Hospital VisitsTell us About the Value of Public Health Insurance as a Safety Netfor the Young and Privately Insured?What the prior literature does not tell usWhat my data can tell us – stylized factsHow I incorporate my data into a framework for valuing public health insurance
What we learn from the framework, along with robustness
The longitudinal value of health insurance is much greater than the sum of its cross-sectional parts
Slide24Appendix Slides
Slide25Hospital Count – SPARCS vs. AHA
Slide26Inpatient Cost – SPARCS vs. MEPS
Slide27Insurance Coverage – SPARCS vs. CPS
Slide28Indifference condition with private information:
Slide29Closed form solution for λ:
Slide30Robustness to Including Uninsured (cont.)Theoretically possible for the longitudinal value of insurance ρ to be equal to zero, even if the risk neutral value of insurance λ is positive