Discussion Paper: Conceptual difficulties in measuring retirement Geoff Bowlby Chief, Analysis and Special Projects, Labour Force Survey Statistics Canada Prepared for the Session on the Aging of the Workforce, Groupe de Paris, Budapest, Hungary May 15-17 th, 2006-03-31 Introduction Currently sitting at 32 million, the Canadian population is rapidly aging. As in many countries, there was a baby boom in Canada that started in the late 1940s, which was followed by a period of declining fertility. Throughout their history, this baby boom generation has had a major impact in the country, and are now about to exert great influence on the size and nature of the workforce. As the eldest of the baby-boomers turn 60 this year, there are a considerable number of people who are expected to retire in Canada in upcoming years. The inevitable wave of retirement will change the general nature of human activity in this country, as a greater share of the population moves from work towards other activities more typical of retirement. As a result, labour force participation is expected to fall, putting pressure on an already tight labour market in this country. How will the retirement boom impact labour markets? Although far from a sophisticated approach to labour force forecasting, the following is a useful illustration of the impact of the baby boom population, who as they exit the labour force should have a strong downward influence on the labour force participation rate and the unemployment rate. Suppose that the current rates of labour force participation for various age groups 1 stayed the same in the future. As people age, they retire (or die) and rates of labour force participation are much lower for older people than they are for younger people. As a result, as the overall population ages, the share of the population active in the economy should fall, even if at any given age the participation rate stays the same. Using this simple scenario, the labour force participation rate should fall considerably with the aging of the population. The downward pressure actually has already begun, and should 1 The age groups used in this analysis were 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69 and 70+
continue for at least the next twenty five years. From highs around 67.5% in 2004, the labour force participation rate in Canada could be in the 60% range within twenty years, all else held equal. Participation rates in the 60% range would be extremely low for Canada. In fact, participation rates have not been that low since the early 1970s, when women were entering the labour force, exerting much upward pressure on the size of the labour supply. Chart 1: Actual and forecast participation rates, Canada, annual averages 2000 to 2017 Participation rate 68 Scenario 2 67 66 65 Scenario 1 64 63 62 Actual Projected 61 60 59 Scenario 1: Medium population growth, participation by age group constant at 2005 rates Scenario 2: Medium population growth, participation among older people rising for next five years then constant 58 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Source: Labour Force Survey, using Statistics Canada population projections Whether or not the aging boomer population actually pushes the relative supply of labour to these lows remains to be seen. A variety of reasons could keep the aging boomers in the labour market longer than earlier generations. Indeed, the Labour Force Survey in Canada has been showing a slow upward movement in the retirement age in recent years, a switch from the longterm downward trend. On the other hand, they may exit sooner than retirees in the past. Regardless of the situation, there are a large number of employed baby boomers (over 7.5 million) whose inevitable exit from the labour market should put downward pressure on the size of the labour supply. What sort of impact would that have on unemployment rates, the most popular indicator of the state of the labour market? That would depend upon what was happening with labour demand,
which has more to do with aggregate economic conditions and technological change than demographics and is therefore difficult to predict. If the demand for labour were to rise, even marginally, in this upcoming era of labour supply contraction, then unemployment rates could fall. In whatever labour demand scenario that eventually comes to fruition, the unemployed should benefit from the eventual retirement of thousands of Canadians and the job opportunities that this will provide. As a result, the aging of the population should exert downward pressure on the unemployment rate. Indeed, in 2005 in Canada, as the relative size of the labour supply contracted (in part because of the aging workforce), and as demand boomed, the unemployment rate hit record lows. So why don t we have a good measure of retirement? The above is not designed to be a rigorous forecasting of future labour market trends. Rather, it is designed to show the potential impact of the eventual retirement wave that will take place. Although there are significant labour market consequences of the wave of retirement that should hit the country over the next twenty years, there is no regular production of statistics on the retired population in Canada. There are some reasons for this. Only recently, has the need for retirement data grown. Secondly, the concept of retirement is to say the least a fuzzy concept. Retirement can mean many things to many people, and measuring it is difficult for National Statistical Organizations (NSOs) everywhere. Having an international standard available would assist in deciding what data, or range of data, should be produced on the subject. The concept of retirement What is retirement? This a complex question, since retirement is both an event, and a state of being. For example, people may go to someone s retirement party, which would celebrate the event of retirement. Or, you may be entering your retirement, a new state of your life. Measuring the event of retirement might be of lower priority for statistical agencies. When it comes to labour statistics, life event measurement is perhaps not required to the same extent as the measurement of the state of human activity. To elaborate through example, estimates of the number of layoffs or exits from employment (the event) are not typically demanded like estimates of the number of unemployed people (the state of the person, which is a result of the event). So, it might be recommended that any international standard setting on the issue of retirement should focus first on defining and then measuring the people who are engaged in the state of retirement. The question should focus first on who is retired, rather than how many retirements have taken place. With that in mind, how might NSO s measure the number of people who are retired? First of all, it could be recommended that household surveys be the main instrument for determining the bulk of statistics on retirement. Supplementing these household survey statistics will be statistics from administrative records, such as pension or taxation records. Surveys of businesses may provide some important information on human resource preparation for retirement, or information on retirement events, but are otherwise limited by the reality that the retired population will not be working, for the most part.
Conceptual difficulties in measuring the retired What does it mean to be retired? Research indicates a large variety of definitions of the state of retirement 2. Because these definitions can be very broad, or more narrow and precise, the choice of retirement concept has a noticeable impact on the size of the population being measured. For some, it can mean complete labour force withdrawal, while for others the retired population may remain partly or fully active members of the economy. Table 1 is intended to show some of the complications of measuring the retired population. Not intended to be exhaustive, the table shows the myriad of work arrangements from which a person may have exited, showing that in some cases the concept of retirement is more applicable than for others. For example, if someone has left a full-time steady job at older age and is now not working and receiving a pension, he/she is undoubtedly retired. But what about a self-employed farmer, who has scaled back operations in older age, but is still engaged in farming? Is this person retired? Building on a table like this one could be a useful exercise for the Groupe de Paris, should it decide to focus future activities on defining the number of retirees. Table 1: Examples of people who may or may not be retired Group Retired 1 Worked steady, full-time paid (employee) job for long period, Yes stopped working in older age, now not working 2 Worked steady, full-time paid (employee) job for long period,? stopped working at that job in older age, now working part-time at another job 3 Worked steady, full-time paid (employee) job for long period, now? working part-time and/or part week in older age for same employer 4 - Worked as self-employed business owner, business now controlled Yes by or sold to someone else, person now in older age 5 Worked as self-employed business owner, business now controlled No by or sold to someone else, person not in older age 6 Worked as self-employed, still employed in the business, now? working reduced hours, person in older age 7 Worked in a series of shorter term jobs, stopped working in older? age 8 - Never worked or long time since last job, person now in older age? On top of showing the complications of defining retirement, Table 1 is also useful for demonstrating two important facts about measuring the retired population: 1) the population should be of older age; and 2) to be retired is a function of previous activity (but not for everyone). 2 Smeeding, Timothy M., and J. Quinn, Cross-National Patterns of Labor Force Withdrawl, presented at the Fourth International Research Seminar of the Foundation for International Studies on Social Security, June 15-17, 1997, Sigtuna, Sweden.
To elaborate, the retired population should be of older age, but that age is up for discussion. Younger people, by definition, ought not to be considered as retired, since the typical concept of retirement is something that is normally reserved for older people. The second point made above is that the retired are defined by their previous activity, but their current labour market activity can influence whether we would want to consider them as retired. For example, someone who has worked as an employee for a long period and then stopped working in older age, only to never work again, should be counted as retired on the basis of his/her past labour market activity, and the nature of the exit. But what if that person who left his/her long term job through retirement was currently employed? Would NSOs want to consider someone who is currently working as retired? In reality, many people do not become retired overnight. Rather, it is a transition as one moves from more intense labour market activity in the past, towards relative economic inactivity, he/she is moving to retirement. At what point along this gradient we should consider the person to be retired is another way of looking at the complications of defining the concept. Earlier in the gradient, one might be considered semi-retired. How Statistics Canada has measured the retired Regardless of the complications of defining the retired, the other reality is that household survey instruments are always going to have limitations in what they can actually measure. One could have a very precisely defined concept, for example, but never be able to measure it. As a result, it can be very useful to examine what has already been applied. a) The standard definition First of all, there is a standard definition in place at Statistics Canada that says retired refers to a person who is aged 55 and older who is not in the labour force and receives 50% or more of his or her total income from retirement-like sources. Ironically enough, the person who led the research for this standard definition has retired since the standard went into place in the late 1990s. Very little is now known about the background research that went into this definition. The definition might seem reasonable, and has the advantage (perhaps) of being an objective measure, not reliant upon personal perceptions. However, it has one important limitation: it can only be applied using household surveys that have both a labour module, and an income module. As a result, only two surveys at Statistics Canada, the Census and the Survey of Labour and Income Dynamics (SLID) can apply this definition. Other key household surveys, like the Labour Force Survey (LFS) cannot use this definition. While SLID and the Census can be a very important tool for retirement research, there is one very important limitation of using these sources neither are very timely. Timely statistics on the issue of retirement are going to become increasingly important in an era when the wave of retirement may lead to significant labour shortages.
b) Other definitions Transitions to retirement using SLID In a recent compendium of research in Canada on retirement, there was an article on the work to retirement transitions 3 in which retirement is defined as a condition achieved when a person leaves the labour market for good, and receives retirement income (Canadian Pension Plan/Quebec Pension Plan, private pension, etc.). Retirement is deemed to have been achieved when a person has spent at least a year out of the labour market, has received retirement income during that period, and does not return to the labour market before the survey ends. Like the Statistics Canada standard, this definition is an objective measure. This definition is intended for application to a longitudinal survey, although retrospective questions on a crosssectional survey that ask when the person last worked could be used to determine who has been out of the labour market for a year or more. As with the Statistics Canada standard, the definition above requires both labour and income modules, and is subject to the same timeliness limitations. General Social Survey The General Social Survey, an annual survey with rotating topics, asked a series of questions on retirement in 2002. With this survey, a series of questions are used to determine if the person was retired. Three types of respondents get categorized as retired : 1) those whose main activity in the last 12 months was retired ; 2) those whose main activity was something else but who said yes to the question Have you ever retired? ; and 3) those whose main activity was something else, who said No, they have not ever retired, but said yes they had stopped working for one of a variety of retirement-related reasons read to the respondent. Unlike the previous definitions, this definition of retirement is not objective rather it relies upon the respondents perception of their retirement status. Some research using this definition has recently been published by Statistics Canada. According to the 2002 GSS, about 1.8 million Canadians were identified as having retired in the 10 years preceding the survey 4. These retired individuals had worked in the past, and were over the age of 50. Again, the definition of who is retired may seem reasonable. And again, the main limitation of these data is their infrequent production. Although much of the 2002 GSS content will be repeated with the 2007 survey, the rising demand for information on the older population might mean that other potential sources of retirement data ought to be investigated. 3 Deschênes, N., and L. O. Stone, The probability of reaching the state of retirement a longitudinal analysis of variations between men and women in New Frontiers of Research on Retirement, Statistics Canada, Ottawa, March 2006. 4 Schellenberg, G. and M. Turcotte, Autumn 2005. Preparing for retirement, Canadian Social Trends. p. 8 11.
Labour Force Survey retirement measures It might seem natural for a Labour Force Survey to measure the retired population, since all concepts of the state of retirement are defined by a person s past and present labour force status. However, the LFS in Canada does not attempt to measure retirement. In fact, relatively little attention is paid to the older population with the questionnaire. On top of this, there is very little data collected that characterizes the not in the labour force population, since the focus of the survey is the characterization of the employed and unemployed. Although the survey cannot count the number of retired people, it is the key source of data in Canada on the average and median ages at retirement. During the interview, all those currently not working are asked when they last worked. If they worked within the past year, they are asked the main reason they left their last job or business, to which they may respond retired. The month and year of retirement is assumed to be the same as that when the person last worked, and knowing this date and the person s current age, an age of retirement can be estimated. All people who say they have retired under the age of 50 are excluded from the calculation 5. Interestingly, these data show that in Canada the median age at retirement fell from 65 in the mid-1970s, to 60.6 in 1997 when the public sector was offering early retirement incentives to cut payrolls. Since 1997, the median retirement age in Canada has inched back up and was 61.0 in 2005. Finally, although one cannot get a count of the retired population using the current LFS questions, a mapping of the not in the labour force survey population reveals some insight into those in their third stage of life. In this mapping, the not in the labour force can be divided into a number of parts, some which are much more likely to be retired than others. For example, of the almost 8 million people not in the labour force in Canada in 2005, about 3.5 million were over the age of 65. Of the others not in the labour force, another 1.2 million men and women did not want any work and were between the age of 55 and 64. Since 1997, all of the increase in the not in the labour force population can be accounted for by these two older groups. Summary and recommendations In short, there will be added demand for data on retirement in the near future, as the inevitable wave of retirement that will take place has its impact on many facets of Canadian life, including the size of the supply of labour. Measuring the number of retirees should become a focus for NSOs, however complicated the concept of retirement. As a method to keep the discussion rolling, there are a few of recommendations that come from this paper: The measurement of the state of retirement should take higher priority over the measurement of retirement events. Household surveys should provide the best estimate of the retired population. Given the possible use of retirement data in understanding labour shortages, the data should be produced on a timely basis. 5 Gower, D., Summer 1997. Measuring the age of retirement, Perspectives on Labour and Income. p. 11-17
The state of retirement should not necessarily mean that the person is economically inactive. In reality, retirement is a process towards such inactivity. Selecting a definition of retirement is really about saying this person is economically inactive enough to be counted as retired. People of relatively young age should not be considered retired. The Groupe de Paris should continue activities toward the development of a standard definition of retirement.