The Effects of State Medicaid Policies on Savings Patterns
and Long-Term Care Decisions of the Elderly
Abstract
At the beginning of the 21st century
an elderly person who anticipates the need for long-term care (LTC) faces
average annual costs of $41,000 to $56,000 to live in a nursing home, and
between $4,500 and $33,444 to receive home- and community-based services. Since most LTC services are not covered
either by Medicare or by private insurance, elderly persons needing LTC must
rely on their own resources, or, when those are exhausted, rely on Medicaid,
America’s asset-tested government provided insurance program. Due to the high costs of long-term care and
the aging of the population, Medicaid expenditures on LTC for the elderly are
expected to double by 2020 in inflation-adjusted dollars. Policy-makers are debating proposals that
will control the growth of expenditures while still helping those in need of
assistance.
This paper undertakes a
comprehensive analysis of elderly persons’ insurance, long-term care and
savings decisions. An understanding of
how differences in Medicaid programs across states and time affect the demand
for and supply of LTC will help to better predict the effects of future changes
in the Medicaid program structure.
States have considerable flexibility in determining Medicaid policies
such as financial eligibility criteria, subsidies for home- and community-based
services, reimbursement rates to skilled nursing facilities, and restrictions
on the construction of nursing home beds.
Using data from the 1993, 1995,
and 1998 waves of the Asset and Health Dynamics of the Elderly, I evaluate the
influences of health and variation in these state Medicaid policies on choice
of long-term care, insurance, and asset and gift levels. This study also models Medicaid eligibility,
since an elderly person might take actions in order to satisfy eligibility
criteria. To control for the impact of
unobserved heterogeneity in all outcomes, the structural equations of the
empirical model are estimated jointly, allowing for correlation in the error
structure across equations and over time.
Such a formulation is essential in order to obtain statistically
consistent estimates, and hence, better predictors of the effects of Medicaid
policy changes on savings and long-term care behavior of the elderly.
Preliminary results indicate that not controlling for unobserved heterogeneity when modeling these choice variables creates biased estimates that misrepresent behavioral responses to changes in important explanatory variables. Estimating the structural model allows for prediction of behavior under alternative policy scenarios. For example, I simulate the effects of permanent changes in various Medicaid eligibility requirements on the choice of insurance, LTC, and subsequent asset allocation, which in turn affects health and future behavior.