Scholz, Seshadri and Khitatrakun Flashcards
(14 cards)
What are their key findings relating to savings in retirement?
That fewer than 20% of HH have less wealth than their optimal targets, and the wealth deficit of those that undersave are generally small.
What study do they use?
They use the Health and Retirement study to calculate net wealth
What issues do they call out in relation to the earnings data?
That socials security earnings are not available for 23% of respondents in their analysis, and that the records are top coded.
What is the retired persons problem?
What is the equivalence scale they use for HHs
n_j = (a_j + 0.7K_j)^0.5
How do they divide HHs?
They divide them into 6 groups according to marital status, education and number of earners in the HH.
They aslso estimate the medical expense specification for 4 groups of HHs, where persistence paramters for medical shocks tightly cluster between 0.84 and 0.86.
How do they solve for the model?
They discretise the state sapce, solve the optimal decision rule for each HH. Then they calculate the optimal consumption for each period for each HH using data on the observed realisations of earnings.
How much of the cross variation in wealth does the model explain?
86% of the cross variation in wealth
What % seem to have deficits?
15.6%
What is the mean magnitude (conditional on having a deficit)?
$5,260
What did they find in relation to housing wealth?
That once we exclude half of housing equity, 61% of all HHs meet or exceed their wealth targets.
What three features of the model did they call out to account for the fact that many HHs seem to be accumulating significantly more that their optimal LC targets?
- Perceptions of real rates of returns - they assume a low rate of return
- HHs may want to leave bequests
- HHs may have expectations of greater longevity
What did they find in their median regression of “saving adequacy”?
- Sharply increasing postivie relationship between the net work surplus and lifetime earnings
- Illustrates that bequest motives are there
Note that medical expense projections are not fully incorporated and so this could signficantly affect future out of pocket changes
How did their model compare to alternative naive models?
- Modigliani - R squared of 6.5%
- MC - R squared of 45.2%
Noted that the decision rules arisin from the augmented lifecycle model are equally critical in arriving at such close correspondence between model and data.
