Health Care Financing over the Life-Cycle, Universal Medical Vouchers and Welfare Juergen Jung Towson University, Maryland Chung Tran University of New South Wales Baltimore - May 2010 Jung and Tran (TU and UNSW) Life Cycle Health 2010 1 / 39 Introduction Main problems of the U.S. health insurance system 1 Health insurance coverage is low 47 million are uninsured in 2006 (15%) 2 Health expenditure is high 16% of GDP in 2006 close to 20% by 2015 3 Questionable health outcomes? Jung and Tran (TU and UNSW) Life Cycle Health 2010 2 / 39 Introduction Analyzing comprehensive health care reform Medicare Prescription Drug, Improvement, and Modernization Act (2003) Health Savings Accounts Medicare Part D (2006) for prescription drugs Stop imports of generic drugs Restrict Medicare's ability to negotiate drug prices Patient Protection and Aordable Care Act (2010) Health insurance exchanges plus mandate Restrictions on insurance companies Expansion of Medicaid Cuts in Medicare Jung and Tran (TU and UNSW) Life Cycle Health 2010 3 / 39 Introduction Health insurance and expenditure proles (2004/05) 911 71 51 31 01 101 61 51 41 31 21 01 91 101 #$ 21 # 21 31 41 51 61 !" 31 41 71 81 51 61 71 81 %& '() *+,-./01223412256 78*(9) 76 * :;<9) & =76 *>* 7? 6 >& @: () 7@A *7@B>*7? 6 >: 9* @B& @'7:9* ) A ;<>; (: * >;? B& @A ; C* ;=*)6 >*? & < *D A EA? * F Jung and Tran (TU and UNSW) Life Cycle Health 2010 4 / 39 Introduction Objectives and main contributions 1 Develop a modelling framework to analyze comprehensive health care reform: Stochastic dynamic general equilibrium life-cycle 1 2 endogenous health expenditures and insurance choice to model with account for the life-cycle patterns of 1 2 2 health expenditures and insurance take-up rates observed in the data Demonstrate the usefulness of the model by quantifying the short-run and long-run eects of a comprehensive reform with universal medical vouchers incl. transitions and welfare analysis Jung and Tran (TU and UNSW) Life Cycle Health 2010 5 / 39 Introduction Related literature 1 2 3 4 5 Health expenditure over the life cycle Grossman (1972a,1972b), Grossman (2000) Consumption over the life cycle Deaton (1992), Gourinchas and Parker (2002) Fernandez-Villaverde and Krueger(2006,2009) Health expenditure shocks, health insurance and precautionary savings Kotliko (1986), Hubbard, Skinner and Zeldes (1995) Palumbo (1999) and DeNardi, French and Jones (2006) Quantitative macromodels with exogeneous health expenditure shocks Attanasio, Kitao and Violante (2008), Jeske and Kitao (2009) Macromodels with endogenous health expenditures and insurance Suen (2006), Jung and Tran (2008) and Feng (2009) Jung and Tran (TU and UNSW) Life Cycle Health 2010 6 / 39 Introduction Outline 1 Model 2 Calibration 3 Policy experiments 4 Conclusion Jung and Tran (TU and UNSW) Life Cycle Health 2010 7 / 39 The Model MODEL Jung and Tran (TU and UNSW) Life Cycle Health 2010 8 / 39 The Model The Model: Key Features 1 An overlapping generations model with heterogenous agents Sectors: household, rm and government Markets: consumption, labor and capital Agents live for multiple-periods as workers and retirees, and face period mortality shocks and labor productivity shocks Incomplete nancial markets 2 New features from the Grossman literature Health as a durable good: consumption and investment Health depreciation and health shocks Endogenous demand for health care and health insurance Mix of private and public health insurance Jung and Tran (TU and UNSW) Life Cycle Health 2010 9 / 39 The Model Preferences and technology Preferences: u (cj , sj ) Health capital: service ow from health capital health production health shocks sj = s ( h j ) (1) hj = h (mj , hj −1 , εj ) (2) pj (εj , εj −1 ) = Pr (εj |εj −1 , j ) Human capital: accumulation ej = e (hj −1 , j ) for j = {1, ..., J1 } productivity shocks πj (j , j −1 ) Jung and Tran (TU and UNSW) = Pr (j |j −1 , j ) Life Cycle Health 2010 10 / 39 The Model Health insurance arrangements A private health insurance market for workers Two insurance plans: individual and group Group insurance oers provided by employers with a probability ωGI (iGI ,j , iGI ,j −1 ) = Pr (iGI ,j |iGI ,j −1 , income ) Health insurance choice: endogenous inj = 0 : no insurance inj = 1 : individual based insurance inj = 2 : group based insurance (if oered via employer) A public insurance program for retirees (Medicare): no insurance choice Jung and Tran (TU and UNSW) Life Cycle Health 2010 11 / 39 The Model Health expenditures and nancing The total health expenditure: pm m Worker's out of pocket health expenditures: o W (mj ) = pm,noIns m min [pm,Ins mj , γ + ρ (pm,Ins mj − γ)] if inj = 0, if inj = 1, 2 Retiree's out of pocket health expenditures: o R (mj ) = min pm,Med mj , γ Med + ρMed pm,Med mj − γ Med Jung and Tran (TU and UNSW) Life Cycle Health 2010 12 / 39 The Model Households Household problem: Timing of events ũ ũнϭ + a’: asset + h’: health capital + in’: insurance + a: asset + h: health capital + in: insurance Shocks: + : health + : productivity + : group insurance x(j) = {a, h, in , , , } Choices: + c: consumption + m: medical services + a’: savings + in’: insurance choice = {c, m, a’, in’} x(j+1) = {a’, h’, in’, ’, ’, ’} Jung and Tran (TU and UNSW) Shocks: + ’: health + ’: productivity + ’: group insurance Life Cycle Health 2010 13 / 39 The Model Households Worker's dynamic optimization problem V j ( xj ) = max n {cj ,mj , aj +1 ,inj +1 } o u (cj , sj ) + βπj Eεj +1 ,j +1 ,iGI ,j +1 |εj ,j ,iGI ,j [V (xj +1 )] s.t. (1), (2) , 0 ≤ aj +1 , and 1 + τ C cj + (1 + g ) aj +1 + o W (mj ) + 1{inj +1 =1} p (j , h) + 1{inj +1 =2} p = we (hj −1 , j ) + R aj + T Beq + Insprot1 + Insprot2 − Taxj + TjSI where Taxj ỹjW = = τ̃ ỹjW + τ Soc + τ Med we hj −1 , j − 1{in j +1 =2} = we hj −1 , j + raj + rT Beq + Insprot1 + Insprot2 − τ Soc + τ Med we hj −1 , j − 1 {in j +1 =2} TjSI p p − 1{inj +1 =2} p max 0, c + Taxj − we hj −1 , j − R aj + T Beq − InsP1 − InsP2 h Jung and Tran (TU and UNSW) Life Cycle Health 2010 i 14 / 39 The Model Households Retiree's dynamic optimization problem V j ( xj ) = max {cj ,mj , aj +1 } n o u (cj , sj ) + βπj Eεj +1 |εj [Vj +1 (xj +1 )] s.t. (1), (2) , 0 ≤ aj +1 , and cj + (1 + g ) aj +1 + o R (mj ) + pjMed Jung and Tran (TU and UNSW) = R aj + T Beq + TjSoc + TjSI − Taxj Life Cycle Health 2010 15 / 39 The Model Firms and Insurance Firms and insurance companies Firms: max {F (K , L) − qK − wL} , given (q , w ) {K ,L} Insurance companies: (1 + ωins ) = J1 X µ Z h i 1{inj (xj )=1,2} (1 − ρ) max (0, pm,Ins mj (xj ) − γ) d Λ (xj ) j =2 j Z J1 X (1 + r ) µ 1{inj (xj )=1,2} p (j , h) d Λ (xj ) j =1 j Jung and Tran (TU and UNSW) Life Cycle Health 2010 16 / 39 The Model Government Government I Bequests: XJ j =1 Z µj TjBeq (x ) d Λj (x ) = XJ j =1 Z µ̃j aj (x ) d Λj (x ) Social Security: XJ Z TjSoc (x ) d Λj (x ) j =J1 +1 Z XJ 1 = µj τ Soc we (j , hj , ) − 1{inj +1 =2} p d Λj (x ) j =1 Jung and Tran (TU and UNSW) µj Life Cycle Health 2010 17 / 39 The Model Government Government II Medicare: XJ µj Z 1 − ρMed max 0, mj (x ) − γ Med d Λj (x ) j =J1 +1 Z XJ 1 = µj τ Med we (j , hj , ) − 1{inj +1 =2} p d Λj (x ) j =1 Z XJ + µ pjMed d Λj (x ) j =J1 +1 j General government budget: G+ = J X j =1 XJ j =1 Z µj Z µj Jung and Tran (TU and UNSW) TjSI (xj ) d Λ (xj ) Taxj (xj ) d Λ (xj ) + Life Cycle Health XJ j =1 Z µj τ C c (xj ) d Λ (xj ) . 2010 18 / 39 The Model Government A competitive equilibrium Given the transition probability matrices and the exogeneous government policies, a competitive equilibrium is a collection of sequences of distributions of household decisions, aggregate capital stocks of physical and human capital, and market prices such that Agents solve the consumer problem The F.O.Cs of rms hold The budget constraints of insurances companies hold All markets clear All the government programs and the general budget clear The distribution is stationary Jung and Tran (TU and UNSW) Life Cycle Health 2010 19 / 39 Calibration CALIBRATION Jung and Tran (TU and UNSW) Life Cycle Health 2010 20 / 39 Calibration Calibration Preferences: u (cj , sj ) = Health services: cjη sj1−η 1−σ 1−σ sj = hj Health production: hj = φmjξ + 1 − δh,j hj −1 + εj Human capital: ej = e (j )χ (hjθ−1 )1−χ for j = {1, ..., J1 } Markov switching probabilities of the productivity shocks and group insurance oer states are estimated from MEPS 2004-2005 data. Jung and Tran (TU and UNSW) Life Cycle Health 2010 21 / 39 Calibration Calibration Baseline Parameters Demographics: J1 = 9 J2 = 5 n = 1.2% Health Production: φ=1 ξ = 0.32 Preferences: σ = 2.5 Health Productivity: θ = 0 , χ = 0.9 Insurance: γ = 1.7% of median income ρ = 34% γ Med = 6% of elderly's aver health spending ρMed = 30% Exogenous premium growth depending on age and health β =? Technology: α = 0.36 δ = 15% g = 1.5% Depreciation rates of health capital δh,j =? Magnitudes and transition probabilities of health shocks εj =? Jung and Tran (TU and UNSW) Life Cycle Health 2010 22 / 39 Calibration The model and the data [1] Insurance Coverage in % 100 50 Model Data 0 20 30 40 50 60 Age [2] Medical Spending in % of Income 70 80 90 40 50 60 Age [3] Average Assets 70 80 90 80 60 Model Data 40 20 0 20 30 1 0.5 Model Data 0 20 30 40 50 60 70 80 90 Age Jung and Tran (TU and UNSW) Life Cycle Health 2010 23 / 39 Policy Experiments POLICY EXPERIMENTS Jung and Tran (TU and UNSW) Life Cycle Health 2010 24 / 39 Policy Experiments A counterfactual health care reform: Universal Medical Vouchers Motivated by Kotliko (2007) and Emanuel and Fuchs (2007) 1 Government issues medical vouchers to all individuals Vouchers are calculated individually based on the amount of the expected health expenditures for next year (experience rating system) Vouchers are nanced by a consumption, payroll, or lump sum tax 2 3 Individuals purchase private health insurance contracts from insurance companies using the voucher Participating insurance companies have to accept vouchers and oer basic insurance contracts Jung and Tran (TU and UNSW) Life Cycle Health 2010 25 / 39 Policy Experiments Steady State Resutls Aggregate eects of the voucher program fnanced by a payroll tax Capital: K Human capital: L Output: Y Med spending: pm ∗ M Consumption: C Consumption tax: τC Payroll voucher tax: τV Interest rate: R in % Wages: w Voucher Payments % of GDP Benchmark 100.00 100.00 100.00 100.00 100.00 5.155 0.000 3.981 100.000 0.0 Regime 1 - Payroll tax 89.823 100.000 96.210 101.538 97.602 5.676 8.227 4.704 96.210 5.208 Table: Steady state result with health as consumption good only θ = 0. Jung and Tran (TU and UNSW) Life Cycle Health 2010 26 / 39 Policy Experiments Steady State Resutls Negative eciency eects driven by: 1 Savings eect 2 Moral hazard eect 3 Tax eect 4 General equilibrium eect 1 + τ C cj + (1 + g ) aj +1 + o W (mj ) + p (xj ) 1 − τ V we (hj −1 , j ) + R aj + T Beq − Taxj + TjSI + vj − τ LS = Jung and Tran (TU and UNSW) Life Cycle Health 2010 27 / 39 Policy Experiments Steady State Resutls Aggregate eciency eects: four key channels Benchmark 100.00 Regime 1 - Payroll tax 89.823 2. Med spending: pm ∗ M (The moral hazard eect) 100.00 101.538 3. Payroll voucher tax: τV (The tax eect) 0.000 8.227 3.981 100.000 4.704 96.210 1. Capital: K (The saving eect) 4a. Interest rate: R in % 4b. Wages: w (The general equilibrium price eect) Table: Steady state result with health as consumption good only θ = 0. Jung and Tran (TU and UNSW) Life Cycle Health 2010 28 / 39 Policy Experiments Steady State Resutls Aggregate eects with consumption tax Benchmark 100.00 Regime 2 - Cons Tax 98.293 2. Med spending: pm ∗ M (The moral hazard eect) 100.00 107.858 3a. Consumption tax: τC (The tax eect) 5.155 18.049 3b. Payroll voucher tax: τV 0.000 0.000 3.981 100.000 3.988 99.382 1. Capital: K (The savings eect) 4a. Interest rate: R in % 4b. Wages: w (The general equilibrium price eect) Table: Steady state results with health as consumption good only θ = 0. Jung and Tran (TU and UNSW) Life Cycle Health 2010 29 / 39 Policy Experiments Steady State Resutls Health as Consumption and Investment Goods Additional channel: the human capital eect Similar results as in the case without human capital eect Higher health spending produces higher health Higher health capital alleviates some of the negative tax distortions Negative savings eect is now smaller Price (G.E.) eects are smaller Jung and Tran (TU and UNSW) Life Cycle Health 2010 30 / 39 Policy Experiments Steady State Resutls Health as consumption and investment good 1. Capital: K (The saving eect) 2. Med spending: pm ∗ M (The moral hazard eect) 3. Payroll voucher tax: τV (The tax eect) 4a. Interest rate: R in % 4b. Wages: w (The G.E price eect) 5. Human capital L (The human capital eect) Benchmark θ=0 100.00 Payroll tax θ=0 89.823 Benchmark θ=1 100.00 Payroll tax θ=1 90.902 100.00 101.538 100.00 101.077 0.000 8.227 0.000 8.224 3.981 100.000 4.704 96.210 4.251 100.000 4.919 96.555 100.000 100.000 100.000 100.199 Table: Steady state results with health productivity θ = 1. Jung and Tran (TU and UNSW) Life Cycle Health 2010 31 / 39 Policy Experiments Steady State Resutls Health as consumption and investment good Previous 4 channels Additional channel: the human capital eect Benchmark 100.000 Payroll tax 90.902 Consumption tax 100.078 Human capital: L 100.000 100.199 100.382 Output: Y Medical spending: pm ∗ M Consumption: C Interest rate: R in % Wages: w Voucher Payments in % of GDP 100.000 100.000 100.000 4.251 100.000 0.000 96.747 101.077 98.346 4.919 96.555 5.227 100.273 107.454 96.548 4.121 99.891 5.673 Capital: K Table: Steady state results with health productivity θ = 1. Jung and Tran (TU and UNSW) Life Cycle Health 2010 32 / 39 Policy Experiments Steady State Resutls Transitions from benchmark to regime 1: payroll tax Output Capital 100 θ=0 θ=1 98 96 100 0 10 20 Time Consumption 30 90 40 80 105 110 100 105 95 0 10 20 30 Time Medical Expenditure 40 120 100 0 10 20 30 Time Health Capital 40 0 10 20 30 Time Human Capital 40 0 10 20 Time Wages 30 40 0 10 20 Time 30 40 100.5 100 % 80 0 10 20 Time Interest 30 40 100 5 100 4 98 3 0 10 Jung and Tran 20 30 Time Consumption Tax (TU and UNSW) 40 96 Life Cycle Health 2010 33 / 39 Policy Experiments Welfare Analysis Welfare eects: Payroll vs. consumption tax Negative eciency eects Capital drops Output drops Household income drops (w decrease) Consumption of C drops Positive insurance eects More insured, improved risk sharing Increases in medical spending Increases in health capital H If H is productive, it has a positive eect on output Jung and Tran (TU and UNSW) Life Cycle Health 2010 34 / 39 Policy Experiments Welfare Analysis Welfare eects: Payroll vs. consumption tax Regime 1 (τ ): Compensating Consumption per GDP (in %) L 6 4 Losses % 2 0 Gains −2 θ=0 θ=1 −4 −6 0 5 10 15 Time 20 25 30 Regime 1 (τ ): Compensating Consumption per GDP (in %) C 6 4 Losses % 2 0 Gains −2 θ=0 θ=1 −4 −6 0 5 Jung and Tran (TU and UNSW) 10 15 Time Life Cycle Health 20 25 30 2010 35 / 39 Policy Experiments Welfare Analysis Aggregate eects: Payroll vs. consumption tax Benchmark Capital: K Human capital: L Output: Y Med spending: pm ∗ M Consumption: C Consumption tax: τC Payroll voucher tax: τV Interest rate: R in % Wages: w Voucher Payments % of GDP 100.00 100.00 100.00 100.00 Regime 1 Payroll tax 89.823 100.000 96.210 101.538 Regime 2 Consumption Tax 98.293 100.000 99.382 107.858 100.000 97.602 96.227 5.155 0.000 3.981 100.000 0.0 5.676 8.227 4.704 96.210 5.208 18.049 0.000 3.988 99.382 5.583 Table: Steady state result with health as consumption good only θ = 0. Jung and Tran (TU and UNSW) Life Cycle Health 2010 36 / 39 Policy Experiments Welfare Analysis Key Lessons General equilibrium channels Tax nancing instruments: Payroll vs. consumption tax Role of health: Productive vs. non-productive health Policy outcomes: Aggregate vs. welfare eects Jung and Tran (TU and UNSW) Life Cycle Health 2010 37 / 39 Conclusion Conclusion 1 2 3 Construct a heterogeneous agents macro-model with health as a durable good Account for life-cycle patterns of health expenditures and private insurance take up rates Assess the macroeconomic eects of introducing a universal health insurance system Jung and Tran (TU and UNSW) Life Cycle Health 2010 38 / 39 Conclusion Future work on macro-health economics 1 Model: Elastic labor and Medicaid The life cycle proles with a pure age eect A structural estimation of the health production function 2 Research Questions: Life cycle consumption puzzle The tax deductible policy and health insurance markets The macroeconomic eects of Obama's health care reform 2010 Optimal public health insurance policy Financing health costs in an aging economy Jung and Tran (TU and UNSW) Life Cycle Health 2010 39 / 39