Directorates
·	Director General's Office
·	Data Development
·	Strategic Evaluations
·	Partnership and Continuous Evaluations
Publications
·	Evaluation Reports
·	Evaluation Briefs
·	Labour Market Development Agreement Series
·	Lessons Learned Series
·	Evaluation Tool Kit Series
·	Evaluation Forum Newsletter
·	From Theory To Practice

A theoretical framework for analyzing the determinants of the retirement decision should serve a variety of purposes. First and foremost, it should provide a comprehensive list of the determinants of the retirement decision and indicate their expected effect on that decision. It should indicate the appropriate functional form or way to enter those determinants (e.g., linear or non-linear, key interactions) as explanatory variables in, say, a multiple regression equation.

A theoretical framework should highlight how to incorporate such factors as life-cycle and family decision-making as well as institutional features such as mandatory retirement and public and private pensions. It should provide guidance as to the appropriate form of the dependent variable (i.e., the retirement decision) as well as how best to estimate the relationship between that decision and its determinants. Ideally, the theoretical framework will be linked to key policy concepts and issues, so that resultant analyzes can be easily linked to policy choices.

Concepts of Retirement: The literature on retirement decisions increasingly recognizes that retirement is not simply a discrete event of leaving the labour force, never to return.¹ It often involves transitions or bridges into retirement, including partial retirement, to use phrases that are increasingly used in the literature. This often occurs in such forms as hours reductions, changing jobs, or intermittent retirement, often involving non-standard employment such as through self-employment, contract work and part-time work.

While most studies tend to use one measure of retirement, some use a variety of alternative measures.² The following concepts of retirement have been used in the literature³:

1. Individual's statement of their planned retirement age.⁴

2. Self-reported response where the person indicates that they are retired.⁵

3. Left the labour force in the sense of no longer working or looking for work.⁶

4. Reduced hours of work or pay, sometimes below a specific fraction of a previous norm.⁷

5. Left career or main employer, possibly to continue working in another job.⁸

6. In receipt of an employer-sponsored pension.⁹

7. In receipt of a public pension.¹⁰

Income-Leisure Choice Framework and Reservation Wages: The theoretical framework that is most often used to analyze the determinants of the retirement decision is the "income-leisure" choice perspective of economics, modified to account for the peculiarities of the retirement decision.¹¹ That perspective views individuals as choosing between retiring from paid labour market activities (i.e., engaging in leisure activities or household work) and continuing in paid employment (i.e., earning income). That decision essentially involves comparing utility or well-being in the alternative states, subject to various constraints such as those imposed by the individual's wealth and the market wage they can expect to receive.

The decision to retire can also be affected by institutional constraints such as mandatory retirement rules, and the incentive effects of public and private pensions (analyzed in more detail subsequently). The retirement decision can also be affected by demand side factors (e.g., labour market conditions and unemployment) that determine whether jobs are available.

An alternative formulation of the decision rule is that the individual will retire if their market wage is below their reservation wage.¹² Their reservation wage is essentially the implied value of their time in non-labour market activities such as retirement. It is higher, for example, if they have higher non-labour market income (e.g., assets, wealth, pension income) from which to enjoy retirement. This can include wealth effects from any intergenerational transfers that may be embedded in pay-as-you go public pension schemes such as CPP/QPP. An individual's reservation wage is also higher if their health or the nature of their work makes it difficult for them to engage in labour market activities. It is also higher if their family circumstances make retirement more attractive (e.g., if their spouse is retired or not working).

The individual's market wage should be broadly construed to reflect the monetary returns to continued labour market employment. Those monetary returns are net of taxes and any transfer payments that may be affected by continued employment. Importantly, they should also reflect the expected public or private pension benefit accruals, or changes in expected pension wealth associated with continued labour market activity. The expected market wage should also be adjusted for demand side factors; that is, the probability of being able to retain or obtain a job.

Life-Cycle Dimensions: Building upon the earlier models that utilized the "income-leisure" choice perspective and reservations wages, subsequent theoretical work tended to emphasize the dynamic life-cycle nature of the retirement decision.¹³ Such multi-period models often involved forward-looking calculations of individuals choosing an optimal retirement age. That optimal retirement age depends upon how they expect their utility associated with retirement versus continued labour force participation to change as they age, as well as how they expect their wages, wealth, health, and pension wealth to change. The life-cycle models emphasize that the retirement decision is not based simply on an evaluation of the different states at a given point in time. Rather, it is based on a more forward-looking approach that evaluates the expected remaining lifetime well-being associated with the alternative states, including the income streams associated with those states.

Re-evaluations and "re-optimization" can be allowed as individuals age and acquire new information.¹⁴ As well, the life-cycle models allow individuals to engage in inter-temporal substitution of labour supply over their life-cycle in response to the changing incentives of the labour market and pension plans.¹⁵

Pension Wealth, Pension Capital Changes and Option Values: The life-cycle framework is also the basis for models that emphasize the importance of comparing the present value of streams of expected pension wealth¹⁶ associated with the decision to retire or to continue to participate in the labour market. It is the present value of those alternative streams that matters, not just their current value at the point in time when the retirement decision may be made.

Changes in pension wealth have been calculated to illustrate this aspect of the incentives that employees could expect to experience if they retired from their particular job.¹⁷ This is often calculated as the difference between their expected "stay" pension wealth if they were to remain in their job until the age of normal retirement, and their "quit" pension wealth if they were to leave their job. Pension capital losses can occur because employees who leave their job forgot the wage increases that would otherwise augment their pension wealth.

Option value measures¹⁸ have also been used to provide a measure of the changing value of pension wealth associated with retirement versus continued employment. The option value captures the notion that a person who retires forgoes the opportunity to continue to work and accumulate additional pension benefit accruals from such factors as additional service credits, wage increases, and eligibility for early retirement provisions. Conversely, continuing to work one more year preserves the option of continuing to work another year, and so forth. This in turn preserves the option of qualifying for such factors as subsidized early retirement provisions. The option value of working each additional year is calculated as the difference in the present value of maximum pension benefits associated with working in that given year as opposed to retiring. Other studies have focused on the exogenous, unanticipated changes in Social Security wealth that has often occurred because of legislated changes in coverage and generosity.¹⁹ Such changes are likely to facilitate retirement (i.e., reduce labour force participation) in large part by enabling individuals to better afford retirement.

Public Pension Plan Incentives: Much of the recent literature on the determinants of retirement decision is of U.S. origin and has focused on the incentive effects of public pensions, notably Social Security in the U.S. Of particular importance is the retirement test which essentially involves reductions or "clawbacks" of pension income for persons above a certain level of earnings who continue to earn income after reaching the age of entitlement for such pension benefits. These are likely to encourage retirement because they reduce the monetary returns to work. Other features, however, can encourage continued labour force participation, for example, so as to continue making contributions that will enhance subsequent benefits.²⁰

In fact, the monetary incentives of public pension plans such as Social Security can be complex given the different rules and requirements of such programs. A number of studies²¹ have calculated and analyzed the changes in expected pension wealth that occur at specific ages when those rules change.

Related regulations give rise to substantial abrupt changes (often termed spikes, kinks, non-linearities, notches) in Social Security wealth at particular ages of individuals. These can substantially affect the incentive to retire since the monetary incentive to work would be high at ages when there is a positive spike in expected pension wealth.

The asset or wealth value of public pension plans can have a different impact on the retirement decision than wealth from other sources. Wealth from public pension plans may have a strong influence in the retirement decision because it cannot be bequeathed to heirs, unlike other forms of wealth which older persons may be reluctant to use up (say, by retiring early) because it can otherwise be passed down as an inheritance (Boskin 1977).

Similarly, public pension income may have a strong effect on retirement decisions because it is received with a high degree of certainty, and often indexed for inflation (Burtless and Moffitt, 1984). Working in the other direction, public pension income may have a weaker effect on the retirement decision because it is not liquid. For example, low-income persons who are liquidity constrained are not able to borrow against future public pension income to retire early.²²

The previous research refers to the effects of U.S. Social Security on the retirement decision. In comparison, evidence on the impact of public pension plans in Canada is extremely scarce. Baker and Benjamin (1997a) analyzed the impact of the removal of the earnings test (clawback of pension payments if the person continued to earn income in the labour market) in the CPP/QPP in the 1970s. Their analysis used the 1972-1980 family files of the Survey of Consumer Finances, conducted every two years. Their results indicated that the removal of the retirement test led to a large and statistically significant increase in weeks worked for those who were employed. Alternatively stated, the clawbacks of the retirement test would have reduced the work time of persons at that time.

Baker and Benjamin (1997b) also analyzed the impact of the early retirement options introduced into the QPP in 1984 and the CPP in 1987. Their analysis was based on the 1972-1980 family files of the Survey of Consumer Finances, conducted every two years. They found little effect of inducing early retirement in the short run, but that in the longer run (after 8 years) the early retirement provisions would have led to a "bunching up" of retirements around those early retirement dates.

Private Employer-Sponsored Occupational Pension Plan Incentives: Significant incentive effects can also be embedded in private, employer-sponsored occupational pension plans. A number of studies have "modeled" or calculated the changes in private pension wealth associated with such factors as the accumulation of age and service credits, early retirement features and postponed retirement features.²³ Such features can give rise to substantial changes in pension wealth in various forms: "backloading" or "deferral" of compensation; spikes in private pension wealth at the ages when early and special retirement features apply; and reductions in pension wealth if individuals postpone retiring past the normal retirement age of their plan.

In most cases, these studies simply modeled the potential incentive effects of the private occupational pension plans by calculating the pension benefit accruals at different ages for persons in representative pension plans. They usually were not able to link these to actual retirement decisions because the data sets on actual retirement decisions did not have detailed information on the private occupational pension plans. Studies that included features of the private occupational pension plans as determinants of the retirement decision are discussed later when the empirical evidence is reviewed.²⁴

These incentive effects of private occupational pension plans can be an important strategic human resource tool for organizations since, in effect, they alter the compensation profile of individual workers. The backloading or deferral of compensation, for example, can reduce unwanted turnover and foster the "bonding" of the employee with the firm (because the employee wants to remain with the firm to receive the deferred compensation or pension benefit accruals). The spikes or large pension benefit accruals at certain ages can encourage early retirement at those ages. The negative pension accruals (i.e., penalties) if the person postpones retirement can discourage postponed retirement and in theory could potentially serve as a substitute for mandatory retirement.

Mandatory Retirement: Mandatory retirement obviously is an institutional rule that can affect — indeed dictate — the retirement decision. Mandatory retirement rules exist as part of company personnel policies or collective agreements, usually as part of employer-sponsored pension plans. They are rules that essentially terminate a particular employment arrangement at a given age. They do not require that the employee leave the labour force, although that may be likely if the alternative employment is unappealing or simply unavailable.

Mandatory retirement rules may exist for a variety of reasons. They may facilitate work sharing by opening job and promotion opportunities for younger workers in the organization. They may facilitate succession planning for the organization and retirement planning and "retirement with dignity" for the individual. They may also facilitate deferred compensation by providing a termination date to implicit or explicit compensation arrangements whereby individuals are "underpaid" (relative to their productivity) when young in return for being "overpaid" when older. Such deferred compensation in turn can serve other positive purposes such as reduced employee turnover and increased bonding with, and commitment to, the firm (Ippolito, 1991; Lazear, 1979).

Whatever their rationale, mandatory retirement can obviously affect retirement decisions especially when they are defined as leaving one's career or long-term job. As such, the existence of a mandatory retirement policy is a relevant explanatory variable to include as a determinant of the retirement decision. Mandatory retirement policies are now banned in the United States. In Canada, however, the Supreme Court ruled in favour of allowing mandatory retirement policies. Such policies tend (but are not required) to apply at age 65 when normal retirement pension benefits are available.

The theoretically expected impact of a mandatory retirement policy on retirement, however, is not as straightforward as it would initially appear. This is so because when mandatory retirement exists an employer-sponsored occupational pension plan is invariably present, and the age of mandatory retirement (e.g., 65) usually corresponds to the age at which public pension plans become available (e.g., CPP/QPP).²⁵

As such, it is extremely difficult if not impossible to disentangle the separate impact of mandatory retirement from the effect of public and private pension plans on the retirement decision. People may leave the labour force around the age at which they are subject to mandatory retirement not so much because of mandatory retirement per se, but because of the monetary incentives to retire as embedded in the associated public and private pension plans.²⁶ But, if mandatory retirement is highly correlated with the onset of public and private pension plans then it is impossible to disentangle the effects of pension features from mandatory retirement rules.

Health, Age, Labour Market Conditions and Other Determinants: The theoretically expected effect of other factors, such as health and labour market conditions, on the retirement decision are fairly straightforward, and studies that use those variables will be discussed later in this report, in the review of the empirical studies.

Ill health likely makes labour market work more difficult and encourages retirement. The retirement-inducing effect of ill health may be particularly strong if individuals also have the income that enables them to afford to retire; that is, health status may interact with other variables to affect the retirement decision.

Labour market conditions, especially the unemployment rate, can affect the retirement decision, albeit in a theoretically indeterminate way. On the one hand, high unemployment may discourage individuals from remaining in the labour market and looking for work.

On the other hand, high unemployment may compel others to remain in the labour market to maintain what otherwise may be declining family income associated with the higher unemployment. As well, these are, respectively, the discouraged and added worker effect²⁷ that higher unemployment can have on the labour force participation (and hence retirement) decision.

Periods of high unemployment, downsizing and mass layoffs can particularly affect older workers because they are often more "expensive" workers and may have difficulty adjusting to the restructuring that is often associated with downsizing.²⁸ If laid off, they may have particular difficulty in finding another job and hence may leave the labour force for retirement.

The composition of jobs and growing wage inequality may also affect retirement decisions.²⁹ The composition of jobs may be shifting towards ones with typically lower retirement ages. Low-wage individuals may not be able to afford to retire, but their lack of viable job opportunities may induce their retirement.

Footnotes

1	Doeringer (1990), Gustman and Steinmeier (1984), Fontana and Frey (1990), Herz (1995), Holden (1998), Honig (1985), Honig and Hanoch (1985), Marshall (1995), Monette (1996), OECD (1995), Parnes and Sommers (1994), Peracchi and Welch (1994), Ruhm (1990, 1991), Swank (1982), and William Mercer (1996).
2	Boskin and Hurd (1978), Gustman and Steinmeier (1984), Palmore, George and Fillenbaum (1982), and Parsons and Lees (1985a).
3	Discussion of different definitions are given in Fields and Mitchell (1984c), Fuchs (1982), Parnes and Less (1985a).
4	Anderson, Burkhauser and Quinn (1986), Luchak (1997).
5	Gower (1997), Hausman and Wise (1985), Quinn (1980, 1981).
6	Anderson and Burkhauser (1985), Anderson, Burkhauser and Quinn (1986), Bazzoli (1985), Burkhauser and Quinn (1983), Frenken (1991), Gordon and Blinder (1980), Hanoch and Honig (1983), Hurd and Boskin (1984), and Quinn (1977).
7	Boskin (1977) used a change to quarter-time work or less, Burtless and Moffitt (1985) used a sudden and discontinuous drop in hours worked, Frenken (1991) used major source of income from pensions, Honig and Hanoch (1985) used half or highest annual earnings as a measure of partial retirement and zero earnings as fully retired, and Reimers and Honig (1989) used monthly earnings below 90 percent of peak lifetime monthly earnings to indicate partial retirement.
8	Bazzoli (1985), Burtless and Hausman (1982), Fields and Mitchell (1984c).
9	Burkhauser (1979).
10	Burkhauser (1979, 1980).
11	Early studies in that genre include Boskin (1977), Boskin and Hurd (1978), and Quinn (1977). Many subsequent studies have built on that framework.
12	Gordon and Blinder (1980).
13	Burbidge and Robb (1980), Burtless (1986), Burtless and Moffitt (1984, 1985, 1986), Crawford and Lilien (1981), Fields and Mitchell (1984a, 1984b, 1984c), Mitchell and Fields (1984, 1985), and Moffitt (1984, 1985, 1986).
14	This is done in the dynamic programming models such as Rust (1987, 1989) and Stock and Wise (1990b).
15	Such inter-temporal substitution of labour supply is emphasized in Burkhauser and Turner (1978, 1981, 1982, 1985), Gohmann and Clark (1989), and McElwain and Swofford (1986).
16	This was first emphasized in Burkhauser (1979, 1980).
17	Allen, Clark and McDermed (1988, 1993), Blinder, Gordon and Wise (1980, 1981), Burkhauser and Turner (1981), Burkhauser and Quinn (1983a, 1983b), Gordon and Blinder (1980), Ippolito (1985, 1986, 1987, 1991), and Kahn (1988).
18	Cornwell, Dorsey and Mehrzad (1991), Lazear (1990), Lazear and Moore (1988), Lumsdaine, Stock and Wise (1990), Pesando, Hyatt and Gunderson (1992), Stock and Wise (1990a, 1990b).
19	Anderson, Burkhauser and Quinn (1986), Burtless (1986), Hausman and Wise (1985), Ippolito (1990), and Moffitt (1987).
20	These are emphasized in Anderson, Gustman and Steinmeier (1997), Blinder, Gordon and Wise (1980), and Gordon and Blinder (1980).
21	Anderson, Gustman and Steinmeier (1997), Blau (1994), Burtless (1986), Burtless and Moffitt (1984, 1985, 1986), Diamond and Gruber (1997), Gustman and Steinmeier (1991), Kahn (1988), Kruger and Pischek (1992), Lumsdaine and Wise (1994), and Moffitt (1984, 1987). Gruber (1997) documents the potential incentive effects from CPP/QPP and other retirement income programs in Canada, based on methodology outlined in Diamond and Gruber (1997).
22	Burtless and Moffitt (1984), Kahn (1988), Nalebuff and Zeckhauser (1985), and Robb and Burbidge (1989).
23	In the U.S. such studies include Allen and Clark (1986), Burkhauser (1979), Fields and Mitchell (1984a, 1984b, 1984c), Gustman and Steinmeier (1989a), Ippolito (1986, 1989, 1990), Kotlikoff and Wise (1985, 1987, 1987, 1989), Lazear (1983), Hogarth (1988), Mitchell and Fields (1984, 1985), Mitchell and Luzadis (1988), and Pozzebon and Mitchell (1989). In Canada, they include Pesando and Gunderson (1988, 1991), Pesando, Gunderson and McLaren (1991), Pesando, Gunderson and Shun (1992), and Pesando, Hyatt and Gunderson (1992).
24	Allen, Clark and McDermed (1988, 1993), Burkhauser (1979), Fields and Mitchell (1992, 1984a, 1984b, 1984c), Kotlikoff and Wise (1985, 1987, 1989), Lazear (1983), Hogarth (1988), Mitchell and Fields (1984, 1985) for the United States, and Luchak (1997), and Pesando, Hyatt and Gunderson (1992) for Canada.
25	A discounted CPP/QPP can be obtained as early as age 60.
26	Burkhauser and Quinn (1983).
27	This is the inducement to increased labour force participation by another family member.
28	Davidson, Worrell and Fox (1996), and Hutchens (1988).
29	Peracchi and Welch (1994).