Understanding the effects of uncertainty about the timing of retirement

Frank Caliendo, Maria Casanova, Aspen Gorry, Sita Nataraj Slavov 16 November 2016

a

A

It is often said that when it comes to retirement, planning is everything. Planning for retirement would be easier if individuals know the exact date when they will retire, as this will determine both the number of years of wage earnings and the expected length of the retirement period. And yet, young people cannot know for certain their future retirement age, as it depends on numerous life events that are difficult to predict in advance. These potential events include changes in family structure and the timing of health shocks or unemployment shocks. Despite the importance of this uncertainty, researchers have paid relatively little attention to its effects on individual decision-making. Several studies have found that unanticipated retirements are associated with a drop in consumption spending at retirement, as individuals who retire earlier than anticipated find themselves with insufficient financial resources (Blau 2009, Grouchulski and Zhang 2013).

A primary objective of social insurance programmes such as Social Security in the US is to prevent poverty in old age. In a world where individuals are uncertain about their retirement date, the risk of entering retirement with inadequate financial resources — and hence of facing poverty — is greater than in an environment where they have knowledge of their retirement date and can make optimal savings plans. A natural question is whether existing social insurance programmes help mitigate the risk from retirement timing uncertainty. To answer this question, in a recent paper we do three things (Caliendo et al. 2016):

  1. We quantify the extent of individual uncertainty about the date of retirement;
  2. We assess how costly this uncertainty is for individuals as they attempt to make optimal saving plans for retirement;
  3. We evaluate how well current government policies such as Social Security retirement and disability insurance help mitigate the cost of this uncertainty.

The extent of retirement timing uncertainty

To understand the effects of retirement timing uncertainty, it is important to first measure how much uncertainty individuals face. We measure the degree of retirement timing uncertainty using data from the Health and Retirement Study (HRS). The HRS is useful for this because when individuals enter the sample in their early 50s, they are asked about the date at which they expect to retire. Individuals are then followed for up to 20 years so that we can observe the date at which they actually retire.

We estimate the degree of retirement uncertainty by computing the standard deviation of the difference between individuals’ expected and realised retirement dates. This reduced-form approach captures both individuals’ retirement expectations and their endogenous responses to shocks that cause them to update their retirement date. Depending on the particular sample used, our estimates of the standard deviation range between four and seven years. This result suggests that even individuals in their 50s, who are well into their working careers, face a great deal of uncertainty about their eventual retirement date.

The welfare cost of retirement timing uncertainty

The next step in our analysis is to measure the cost of this uncertainty. At an intuitive level, if an individual who expected to retire at 65 receives a shock that induces him to retire at age 60 instead — approximately one standard deviation prior to his expected retirement — he would lose five years of lifetime wage income, and would need to finance an additional five years of retirement. With this idea in mind, it is possible to formulate a quantitative life-cycle model of individual consumption and savings to compute the welfare cost of this uncertainty. We do this by using our reduced-form measure of retirement uncertainty, to calibrate the distribution of retirement uncertainty in the model.

There are a number of alternative ways to measure the welfare cost of retirement uncertainty. First, we can compute the percentage of lifetime income that individuals would give up in order to live in a world with no retirement uncertainty in which they are endowed with lifetime wealth that is the same as their expected lifetime wealth as in the world with uncertainty. This measure provides the value of having full insurance with respect to the variation in wealth across different realisations of their retirement. We find that individuals would give up 2.6-5.7% of their lifetime income to live in the riskless world.

A second useful measure of the cost of retirement uncertainty is the fraction of lifetime income that individuals would give up just to know the date of retirement at the beginning of their work life. In this case, the realised date of retirement determines lifetime income, but the individual can make optimal savings plans with knowledge of when they will eventually retire. Using this measure, we find that individuals would give up between 1.9% and 4% of their lifetime income depending on the standard deviation of retirement timing. Since these costs are nearly as large as the full insurance costs, we conclude that much of the welfare cost is due to distortions in consumption and savings plans, from not knowing the date of retirement. In other words, simply having information about the timing of retirement is almost as valuable as protecting one’s wealth from retirement timing risk.

Finally, we can also compute the costs of retirement uncertainty in a model where individuals learn their date of retirement at a particular age before retirement, rather than at the date when the shock forces them to retire. We make the conservative assumption that individuals learn their actual retirement date at age 50, earlier than we measure their retirement uncertainty in the data. In this case, the welfare cost to individuals of not knowing their retirement date is still 1.5% of lifetime income, putting a lower bound on the welfare costs that individuals face. To put these values in perspective, 1.5% of lifetime income is larger than either the cost of business cycles or fluctuations in wage income as measured by Lucas (2003) or Vidangos (2009), respectively.

The role of social insurance programmes

We use our quantitative model to determine how well social insurance programmes such as Social Security’s retirement and disability insurance protect individuals against the risk of retirement timing uncertainty. The answer is that these programmes do little to mitigate retirement timing risk. To understand why, it is helpful to think about the structure of a first-best insurance policy against retirement uncertainty. To insure against timing risk, such a policy would need to pay individuals a large amount if they unexpectedly retire early, and a small amount if they retire late.

Some features of the Social Security retirement programme can provide partial insurance against early retirement shocks, as individuals who retire early contribute less in Social Security taxes and receive a higher return on their contributions through the progressivity of the benefit formula. But these features are more than offset by the fact that those who suffer early retirement shocks have their number of working years cut short, and will generally have lower average earnings. On the whole, unlike the first-best benefit schedule, Social Security retirement benefits tend to be lower for those who retire earlier than expected and higher for those who retire late, making the programme ineffective at providing timing insurance.

To study the impact of the Social Security disability insurance programme, we extend the model to include the potential for disability-induced retirement. In that case, we find that the disability insurance programme does a good job of offsetting the additional disability risk that individuals face, but does not do very much to insure individuals against shocks to the timing of retirement. This is because the programme does not grant higher benefits to those individuals who are hit by a disability shock earlier in their careers, and hence once more looks nothing like the first-best benefit schedule.

Conclusion

We find that individuals face a large amount of uncertainty about when they will eventually retire, and that this uncertainty is costly to them because it makes it difficult to appropriately save for retirement. Moreover, the current structure of the Social Security retirement and disability programmes does not provide much insurance against this risk. Given these challenges, it is worth trying to understand what changes to social insurance programmes could help reduce this important risk. One possibility is to make the Social Security retirement formula less dependent on individual income, as is done in Japan and many European countries where part of the retirement benefit is independent of past earnings. But while such changes can help insure individuals against retirement timing risk, they may change labour supply incentives. These tradeoffs should be further explored.

References

Blau, D M (2008), “Retirement and Consumption in a Life Cycle Model”, Journal of Labor Economics, 26, 35-71.

Caliendo, F N, M Casanova, A Gorry, and S Slavov (2016), “The Welfare Cost of Retirement Uncertainty”, NBER Working Paper no. 22609.

Grochulski, B, and Y Zhang (2013), “Saving for Retirement with Job Loss Risk”, Economic Quarterly, 99, 45-81.

Lucas, R E (2003), “Macroeconomic Priorities”, American Economic Review, 93, 1-14.

Vidangos, I (2009), “Household Welfare, Precautionary Savings, and Social Insurance under Multiple Sources of Risk”, Federal Reserve Board Working Paper.

a

A

Topics:  Labour markets Welfare state and social Europe

Tags:  Retirement, work, Labour Markets, uncertainty, insurance, Social security

Head of the Department of Economics and Finance, Jon M. Huntsman School of Business, Utah State University

Assistant Professor, Mihaylo College of Business and Economics, California State University, Fullerton

Assistant Professor of Economics and Finance, Utah State University

Professor of Public Policy, Schar School of Policy and Government, George Mason University

Events