Bernheim et al. Flashcards
(14 cards)
What surveys do they use?
Uses the Panel Study of Income Dynamics and the CEX.
How do they account for the variation in wealth, fixing the income profiles (helps to focus on the earnings replacement rate)?
(1) Slope of the consumption profile - should have positive correlation (dependent on preferences etc.)
(2) drop in consumption at retirment - should have negative correlation with retirement wealth
(3) Overall level of consumption - bequest motives for higher wealth
LC model predicts that, holding preferences constant, HHs facing lower replacement rates should save more - and so controlling for income profiles, the authors isolate the impact of preferences and other factors on wealth accumulation.
How do they account for the variation in wealth that results from differences in income profiles?
HH saving should vary inversely with predicatable differences in earning replacement rates.
Tastes for savings should not be correlated with the size of decline at retirement - but this would be modified if the taste for thrift also relates to the magnitude of the consumption discontinuty at retirement.
How did they restrict the sample?
They restricted the sample to HH with non lenghty transition periods (more than two years in transition)
Did they do anything regarding the unbalanced nature of the panel?
They found that it did not matter for their results, found similar results using subsamples with balanced panels.
Why did they complement the PSID data with CEX data?
PSID does not have ideal data on consumption - using the CEX helped to increase the predictive power of the PSID threefold.
What do the quartiles in their data correspond to?
Divided sample into 4 equally sized quartiles based on the ratio of non asset income for the first three years of postretirement to non asset income in the last three years before retirement.
Fourth quartile - includes high income HHs with generous post retirement packages, as well as low income with high social security benefits.
Do they account for financial wealth vs total wealth?
Yes - they use specifications for both and find that this doesn’t shift the results considerably.
What was their key result?
The found that there were large discontinties in retirement, and that there are differences in this magnitiude as it relates to wealth quartiles.
Similar patterns for the income quartile.
The magnitude of the discontinuity declines monotonically with the HH’s wealth ratio quartile and with its income replacement quartile.
How did they remove the news associated with retirement?
Exploits the fact that the retirement hazard function varies sharply with age (spikes at 62 and 65)
What did they find regarding “news” associated with retirement?
They fond that even when you remove the effects of unexpected retirement, the size of the consumption discontinuity is still strongly related to wealth and income.
What did they find in their investigation of the composition of consumption?
They found that while retirement was associated with lower budget shares for potentially work related goods, there was relatively little variation by wealth in the size of these declines for categoriees like clothing and fuel.
Used CEX data and an conditional commodity demand approach.
Not enough to explain the variation in wealth across the population.
What consumption dynamics are they looking to estimate for.
Estimate functions relating changes in log consumption to time (relative to retirement), household fixed characteristics, and demographic factors that may change over time (e.g., household size).
Specifically, examine the correlation between accumulated wealth and:
- The average consumption growth rate (slope of the consumption profile).
- The size of the drop in consumption at retirement (discontinuity).
Do these results support “rule of thumb” or hyperbolic discounting theories of wealth accumulation?
Yes, the data was consistent with this conclusion.
The consumption patterns that they observed did not match what economic theoru predicted for HHs with different wealth accumulation rates.
