The Effect of Tax Preferences on Health Spending John F. Cogan R. Glenn Hubbard Daniel P. Kessler Hoover Institution Columbia University Hoover Institution Stanford University NBER Stanford University NBER Taxes and Health Insurance Conference Washington, DC February 29, 2008
Overview: Tax Preferences for Health Care • Employer-sponsored health insurance is deductible to employers but excludable from employees’ income and Social Security tax base • Budget cost larger than mortgage interest, state and local tax deduction, all capital gains preferences • Distortion of consumption choices • Too much health spending • Too much insurance for any given level of spending • What is the empirical magnitude of the effect of the tax preference on health spending?
Identification problem Income Marginal tax rate How can observational data distinguish between these two effects? Post-tax price of insurance Health spending
Identification strategy • Consider a sample of individuals with full-year employer-sponsored health insurance obtained through a single full-year worker • Compare the health spending of individuals in families with a policyholder who earns just above the Social Security threshold to spending of individuals in families with a policyholder who earns just below it • Difference in their health spending, conditional on other covariates, captures the effect of the tax preference
Data Medical Expenditure Panel Surveys from 1996-2005 • For each individual, survey provides amount of spending on health services, wage earnings, other demographic and economic information • Survey also links each individual who has employer-sponsored insurance to the worker who holds the policy that is the source of coverage
Effect of the tax preference: IV model Characteristics of individuals and families Health spending of person i, family j, year t Error terms After-tax price of health insurance Function of family income and salary of insurance policyholder g() includes ln(family income) and spline in policyholder’s salary Estimate equation by instrumental variables with indicator for whether family’s policyholder’s salary is 100-110% of the Social Security wage base as excluded instrument
Policy implications • Repeal of all federal tax preferences for health insurance would lead to a 41.6% decrease in health spending by full-year employer-sponsored insured • Many overstate true effect if elasticity of spending with respect to taxes declines as tax preference is phased out • May understate true effect if nationwide change in tax policy led to systemic changes in efficiency of insurance and medical care production processes
Policy implications • Magnitude of effect is consistent with other research • Compare to Gruber (2002) simulations • Removing federal income tax subsidies alone would result in a 35.4% decrease in spending • Removing all federal and state tax subsidies would result in a 50.3% decrease in spending • Compare to Jack and Sheiner (1997) simulations • Removing all federal tax subsidies would result in doubling of effective coinsurance rates • Doubling of coinsurance rates would result in 13-47% decrease in spending (depending on elasticity of demand for health services)