The Effects of Means-Testing Pensions on Savings and Retirement

NIESR Discussion Paper Number 265. Report produced by the National Institute of Social and Economic Research for the Department for Work and Pensions.

Summary
This study uses a stylised simulation model to consider potential behavioural responses to the reduction in pension taper rates associated with replacing the Minimum Income Guarantee (MIG) with the Pension Credit (PC). A range of long-run responses to the policy change are identified, summarised by four behavioural "types":

1. Households on very low incomes, who do not save any wealth under the MIG Ñ comprising 17 percent of the population in 2004. The behaviour of these households is affected only slightly by replacing the MIG with the PC. These households do not save for retirement and choose to exit the labour market before state pensionable age because of their low labour incomes (relative to welfare benefits), rather than in response to the effects of means tested pensions.

2. Households (other than those described under 1) for which a marginal increase in savings would be subject to a 100% taper under the MIG Ñ comprising 12 percent of the population in 2004. Substitution effects dominate with the introduction of the PC, and these households choose to increase their savings for retirement (by £11,000 on average between ages 60 and 64), and to choose a later retirement age (an additional 18% to 20% choose to be employed between ages 60 and 64). Interestingly, the increase in government expenditure implied by the reduced taper rates on private income for this population subgroup under the PC is more than off-set by the increase in revenues associated with the subgroup's increase in savings. The responses of this population group consequently appear to most accurately reflect the suppositions of those who argued for a move away from the MIG.

3. Households (other than those described under either 1 or 2) with sufficient savings under the MIG to generate incomes close to the upper threshold for which benefits can be received under the PC Ñ approximately corresponding to the third population quintile (31% of the population1 ). These households are unambiguously better off under the PC. In the absence of behavioural responses, some of these households would receive a state retirement benefit under the PC where they would not under the MIG, and the PC implies a smaller cost of consumption during the working lifetime than the MIG. Hence, both substitution and income effects motivate these households to reduce their retirement savings (by £6,500-£7,300 on average between ages 60 and 64), and to take earlier retirement (between 4% and 7% fewer choose to be employed between ages 60 and 64) under the PC relative to the MIG. The behavioural responses of these households draw a clear contrast with those described under 2.

4. Households that earn sufficiently high incomes, so that they lie beyond the thresholds of means tested pension benefits under either the MIG or the PC Ñ comprising the fourth, and in particular, highest population quintile (approximately 40% of the population). As could be expected, these households do not appear to be much affected by the policy counterfactual considered here. Aggregating the behavioural effects over a population cohort suggests that substitution effects on poorer households will dominate the long-run impact on retirement of the shift from the MIG to the PC, so that an additional 0.5-1.4% of the population will choose to supply labour between ages 60 and 64. In contrast, the offsetting savings responses of alternative households produce an insubstantial fall in aggregate population savings. In terms of the long-run budgetary impact of the change from the MIG to the PC, the study supports the view that the reduction in pensions means testing will reduce the reliance on the welfare state. The effect of the policy change on the aggregate Government budget, however, suggests overall budget neutrality. Under the PC, the effects of reduced government expenditure on means-tested benefits, reduced tax receipts from savings and increased tax receipts from working longer all offset one another.

This last observation is a key result of the current study, and arises due to an interesting relation between the behavioural responses of lower and middle income households. In the absence of behavioural responses, it is clear that many households in receipt of the MIG would receive a larger state benefit under the PC. Our analysis, however, implies that lower income households will actually impose a smaller burden on the public purse under the PC than they would have under the MIG, because of their responses to the improved incentives to save. This net benefit to the government budget is almost exactly offset in our analysis by the responses of middle income households, who save less and retire earlier under the PC. The finding of budgetary neutrality is important as it suggests that the policy reform is welfare improving, since all households are at least as well off under the PC as they would have been under the MIG, and some are made strictly better off.

Sensitivity of the above results to aspects of the policy environment are also reported here, and are summarised in Table 1. This table indicates that reducing means testing beyond the PC by providing a Citizen's Pension equal to the MIG, exaggerates the behavioural responses of households in the first and second quintiles, increases the consumption (both pre and post retirement) of households above the second quintile, and imposes a higher cost on the public purse.

Furthermore, we consider the effects of allowing the PC to extend up the income distribution, as is implied by the differential indexing that is currently applied to the Savings Credit. This analysis suggests that extending the PC up the income distribution will not have a substantial effect on the behaviour of the lowest quintile (as these households are indifferent to policy changes that affect incomes beyond a lower threshold), will reduce the behavioural responses observed for the second quintile (as the range of means tested incomes in retirement is extended), will motivate the third quintile to save and work less (as a larger proportion are subject to means testing), and motivate households in the fourth and fifth quintiles to save more (due to the income effects of a relative fall in the value of the basic state pension).