De Nardi et al Flashcards
(19 cards)
What is the timing that the author refer to as it relates to medical expenses and survival shocks?
1) Individuals health status and medical expenses are realised
2) Individual consumes and saves
3) Survival shock hits
What is the individual’s value function?
Note that x is cash on hand which is assets minus consumption.
How is health status uncertainty calculated?
Using transition probabilities (logistic regressions). This is dependent on sex, age and permanent income.
Why did they decide to use conditional medians rather than means?
This reflects sample size considerations - when analysing wealth data by such granular groups, some cells only contained a limited number of observations and median is robust to outliers.
How are medical expense shocks simulated?
Using Monte Carlo methods.
What are the two econometric considerations that the authors need to account for?
(1) Cohort Bias - average wages lower for older cohorts due to economic growth.
SOLVE: structural approach
(2) Mortality bias
Wealthier and healthier people tend to live longer. So as people age, the group of survivors become increasing made up with those sorts.
SOLVE: structural - allows mortality to vary with sex, perm income and health status.
What survey data do they use?
The use the AHEAD data as part of the Health and Retirement Survey
What is the AHEAD data used for?
AHEAD data is used to construct the initial distribution of permanent income, age, sex, health and cash on hand that starts off the simulations.
What is one caveat for the AHEAD data?
That it doesn’t include Medicaid expenditures and so the medical expense process they feed into the model is considered a conservative one.
How do they split the asset profiles?
Into 5 birth year cohorts and further split each cohort into 5 perm quintiles
For the medical expense and income profiles, why do they use a fixed effects estimator as opposed to OLS?
Interested in the variation with the same individuals as they grow older; helps to capture cohort effects.
What did they find regarding medical expenses and risk?
That most of the HH lifetime medical expenses comes from the persistent component, after acknowledging that the medical expenses for the elderly are volatile as well as high.
What were the key estimated parameters for the benchmark model?
v (RRA) was 3.8.
Beta - 0.97
Consumption floor - $2,663
Note that a higher fixed floor worsens model fit (and they mention how it is tricky to pin down a specific number due to program complexity and take up)
What is the issue with the bequest motive parameter in the model?
It did not improve the models fit, but point estimates show that the bequest motive becomes operative at £36k, and most individuals in the sample don’t reach this threshold.
What were their findings in relation to the marginal propensity to bequeth?
That the marginal propensity to bequeth is high, where 88% of each dollar above the threshold is left as a bequest for those above the operative threshold.
Implies that bequest motives may be relevant for the richest individuals in the sample.
How did they evaluate the important determinants of savings using the model?
They ruled out attrition so to focus on underlying changes to savings, assuming that every individual lives to 100.
Zeroed out relevant factors to assess their effect.
What was their key finding here regarding out of pocket medical expenses and medical risk?
That when you zero all out of pocket medical expenditures, these are a big determinant of the saving behaviour, particularly for those with high perm income.
The risk associated with the volatility of medical expenses only has a small effect on the profiles of median wealth.
What happened when they lowered the consumption floor by 20%
Most tended to save more as a result and this mattered for all quartiles evaluated.
Why did they look to endogenise medical spending in the model and what did they find?
Used this to check sensitivity of findings - found that the effects on consumption floor are smaller because retirees can adjust medical expenditures as well as consumption - and confirms initial findings that medical expenses are major driver of savings.
