Attansio and Rohwedder Flashcards
(19 cards)
What issues with data sets did they acknowleged to understand savings for retirement?
- Aggregate time series is of limited use due to aggregation issues
- Datasets with individual level data is few and far between and such data sets rarely have information on financial wealth
- Pension wealth does not exhibit much exogenous variation as microdata often covers short periods of time
What is their solve for challenges in measuring pension wealth and what data set do they use?
Using savings rates, because they can use the UK Family Expenditure Survey that computes savings rates as the residuals between disposable income and total expenditure.
This survey is a time series of cross sectional HH surveys.
What method did they use?
Using three major reforms, they used differences-in-differences estimation so to derive an empirical estimation that can be directly interpreted as the degree of substitutability between financial and pension wealth - they then let this degree of substitutability vary with age.
Why did they not analyse data beyond 1988?
This is becuase until 1987, participation in private pension would concievably be assumed to be exogenous with respect to savings choices - and so SERPs introduction did not crowd out pensions provisions.
SERPS was not a choice variable for the worker, but after 1988, they could opt out and so it was no longer an exogenous event.
What were the three reforms they exploited variation for?
(1) 1975 - change in BSP indexing, from ad hoc to average earnings growth
(2) 1979 - introduction of SERPs - pays an additional pension that is linked to earnings (based on best 20 years of earnings)
(3) 1981 - BSP indexing now just linked to prices growth
Why did they look to 4 occupational groups upon analysis of SERPs?
Because it was earnings relation, there exists variation due to differences in individual earnings level.
How was savings measured?
The residual of disposable HH income and total expenditure (which also included spending on durable goods)
Not an ideal measure as it excludes capital gains on real estate and financial assets.
How did they calcuate (subjective) pension wealth?
- Sum of future benefits (assuming continued participation) minus future contributions
- Expressed in constant prices, and assumes perfect foresight about inflation rates
- Accounts for uncertainty about longevity by applying survival probabilities
- Assumed people will retire at state pension age
Why does expecte pension wealth increase over the lifecycle?
(1) benefits closer in time
(2) incorporates survival probabilities
(3) fewer remaining contributions to deduct
What was the impact upon pension wealth as it relates to the reforms?
First indexation change led to a 50% decrease in PW on average, but this varies by cohort type (i.e. youngest lost 80%)
SERPs - benefited younger cohorts PW, but we see larger differences across occupational groups
Outline the LC model that they use
Analyses a 4 period model - so can study how a pension reform affects people of different ages.
- Period 4 is where they retire and use benefits b - the use of the 4 periods
- Savings are assumed to appreciate over time at an exogenous and fixed interest rate of r.
- Do not face liquidity constraints
What is the specification they use for a multiperiod framework, and can you explain what the gamma coefficient means?
They base it on savings, rather than consumption for a better error structure.
The function in front of the EPW is a normalisation factor.
Gamma is a measure of substitutability of financial vs pension wealth.
What relationships do they find with saving/consumption and pension wealth as calcuated?
That an increase in EPW leads to an increase in consumption spending of 80 pounds
Why do they not use OLS?
Would be biased as subjective pension wealth unlikely to be equal to actual, plus they only have an approximation
How do they instrument for their EPW measures to capture the differential impact of pension reforms across the demographic and occupational groups over time?
Uses the interaction of group dummies and year dummies as instruments for their EPW measures
The variation in pension wealth is not fully explained by just group and year dummies; there must be “variation over and above” this, which is provided by the differential impact of the reforms ( they test this using F test)
How do they treat the earnings variable?
They choose to control for earnings using group and age effects, a more robust appraoch than estiamte the PDV of future earnings (as exogenous variation is lacking here).
What were the key results for SERPS for older cohorts?
Pension wealth from SERPS is a strong substitute for private financial savings.
For the oldest cohort, the estimated offset is -0.75, meaning a £1 increase in SERPS pension wealth reduces private savings by 75p
What was the impact on pension wealth SERPs had on younger cohorts?
The substitution effect is not significantly different from zero—no strong evidence that increased pension wealth reduces private saving.
This may be due to liquidity constraints (young people can’t or don’t want to reduce saving now in response to future pension wealth). SERPS has a min accrual.
What was the impact the changes in the BSP indexatino had on private savings
Only affects the youngest cohort, and even then, the effect is small (-0.3). This may reflect that the BSP is less salient or less substitutable for private savings.
