A MACRO-ECONOMIC INDICATOR OF AGE AT RETIREMENT Robert L. Brown,* Robin Damm,† and Ishmael Sharara‡ ABSTRACT This paper explores the relationship between the Wealth Transfer Index (WTI), a statistic deﬁned by Brown and Bilodeau (1997), and retirement age, which is the age at which the workers in an economy cease to be economically productive. Appropriately expressed as a ratio of consumption demand to labor productivity, WTI is a barometer for the demand for wealth placed on the workers of an economy. This paper explains why a relationship between this statistic and retirement age must exist. Using Canadian historical median retirement age data compiled by Statistics Canada and calculated values of the WTI for the same period, three linear regression models are ﬁtted. The conclusion from these models indicates that there is a strong positive correlation between the WTI and average retirement age. This paper also brieﬂy looks at the well-documented demographic shift expected to occur in Canada because of the baby boom– baby bust tidal wave. The aged dependency ratio is expected to increase dramatically, reaching 45% in 2036. A practical application of the WTI model suggests that the baby boom cohort may experience a rise in the normal retirement age in the period 2017–34. They will, in effect, be forced to retire at ages that will allow for an “acceptable” transfer of wealth from the workers to dependent Canadians. Using one of the ﬁtted linear regression models, and projected values of the WTI, the paper concludes by projecting the median retirement age to 2041 for Canadian workers. INTRODUCTION tions, charitable donations, and voluntary altru- In an economy it is the labor force that produces ism. the goods and services demanded by consumers, Brown (1999) argues that there is a limit on the both productive and unproductive. Through demand for wealth transfer that can be borne by large-scale macro-economic wealth transfer the workers in an economy. He maintains that schemes such as social security, education, un- the economy has “self-adjusting” mechanisms employment insurance, and national health care, (acting through the government, employers, indi- wealth is redistributed from the workers to the viduals, or economic forces) that always ensure beneﬁciaries of these schemes. The redistribution that, at any given time, only an “acceptable” de- is accomplished by means of the sale of assets by mand for wealth by dependants can be placed on retirees (e.g., pension asset liquidation), taxes on workers. One way to maintain an acceptable bal- the workers’ earnings, social security contribu- ance is a shift in the average age at which a worker retires (i.e., ceases to be economically productive). Workers will retire at the earliest age *Robert L. Brown, Ph.D. F.S.A., F.C.I.A., A.C.A.S., is Professor in the that is “affordable” in a macro-economic sense. Department of Statistics and Actuarial Science, University of Water- This paper attempts to validate this argument loo, Waterloo ON Canada N2L 3G1, e-mail, email@example.com. by using a statistic called the Wealth Transfer † Robin Damm, MMath, is an Actuarial Associate at Towers Perrin, 150 Index (WTI), which measures the relative de- Sixth Ave. SW, Suite 3700, Calgary, AB T2P 3Y7, Canada, e-mail: mand for wealth placed on the labor force by the firstname.lastname@example.org. ‡ Ishmael Sharara, MMath, A.S.A., is with Hewitt Associates, 25 Shep- young, unemployed, and aged. Using Canadian pard Ave. West, Toronto, ON M2N 6T1, Canada, e-mail: data, we show that a strong positive correlation email@example.com. between the WTI and average retirement age in 1 2 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 5, NUMBER 2 fact does exist, thus supporting the hypothesis as is. First, the study on which the weights are that the average retirement age of a given popu- derived is now 20 years old. It is true that, in lation will shift to allow for a stable wealth trans- 1982, Foot (1982, p. 135) corroborated the fer from the workforce to all dependent Canadi- weights (and suggested that, in the United States, ans. The paper concludes by projecting the the ratio of transfers to the aged would be about average age at retirement for the Canadian work- three times the transfers to youth), but no later ing population for years to 2041, using historical data exist. There are many reasons that, over a retirement age data from Statistics Canada, aver- 20-year time span, the weights would have age rates of productivity improvement, and fore- shifted. Educating the young has become more casted values of the WTI. expensive, as has health care for the elderly. Some social security payments (e.g., Old Age Se- curity and the Guaranteed Income Supplement) THE WEALTH TRANSFER INDEX are indexed to inﬂation, whereas others (e.g., the The WTI, developed by Brown and Bilodeau Canada/Quebec Pension Plans) are indexed (prior (1997), is a statistic that measures the relative to retirement) to wages. Furthermore, ad hoc supply of and demand for wealth among the Ca- amendments to all of these plans have taken place nadian population. It is deﬁned as over this 20-year period. At the end of the day, however, these data are all that are available. 1.866 Y 1 U 4.636 A Thus, we argue that we should continue the study WTI , LF as outlined. where Y Youth, 0 –19 CANADA: DEMOGRAPHICS AND U Unemployed adults IMPLICATIONS FOR THE FUTURE A Aged, 65 and over The Canadian baby boom– baby bust demo- LF Employed labor force ages 20 – 64. graphic proﬁle has been well documented. The The weights of 1.866, 1, and 4.636 were derived rise in birth rates during the 1950s and early by McDonald and Carty (1980, pp. 16 –17) for the 1960s, coupled with the dramatic decline in these Task Force on Retirement Income Policy (1979) rates in the 1970s, will shift the population age and depict relative wealth transfer weights for the structure over the coming years. Exacerbating young, unemployed adults, and the elderly, re- the rise in the number of seniors in Canada is the spectively. The weights do not have any meaning fact that life expectancy is increasing (see Table by themselves; they are only weights relative to a 1). Figure 1 outlines the historical and projected weight of 1 for unemployed adults. It is important distribution of the young, adult, and the aged in to note that the transfers to the aged are almost Canada to 2100. Clearly, this “aging” of the pop- exactly 2.5 times the transfers to youth. These ulation will create a heavy demand for wealth weights are based on total payments for health transfer from the workers to the elderly, which care, education, unemployment transfers, and re- could create pressure for an increase in taxes and tirement income security made by any govern- other contributions from the workers’ earnings, ment (federal, provincial, or municipal). While all else being equal. this does not represent the totality of dependency However, this shift could mean that baby costs, it does capture the key macro-economic boomers simply will not be able to retire at the indicators. It should be noted that a factor for ages currently accepted as the norm. There are productivity improvement should be included in several reasons why this might come true. As- the denominator for comparisons of wealth trans- sume that the massive baby boom cohort at- fers over a period of years. For example, even if tempted to retire at ages now accepted as normal. the demand for goods and services by dependants As the baby boomers attempted to liquidate their were to grow, the increased demand for wealth assets to buy goods and services, these asset transfer could be met if the workforce became prices could become depressed. more productive. Furthermore, because the much smaller baby There are problems with the use of this index bust generation is now the only source of labor, A MACRO-ECONOMIC INDICATOR OF AGE AT RETIREMENT 3 Table 1 Life Expectancy in Canada (1931–1994)* At Birth At Age 65 At Age 75 Year Male Female Male Female Male Female 1931 60.0 62.1 13.0 13.7 7.6 8.0 1951 66.3 70.8 13.3 15.0 7.9 8.8 1971 69.3 76.4 13.7 17.5 8.5 10.7 1991 74.6 80.9 15.7 19.9 9.6 12.5 1994 75.1 81.1 16.1 20.1 9.9 12.7 *These are period life expectancies based on the q’s experienced in the reference year, without projection. Source: Statistics Canada (1986, 1995a, 1997). production in the economy could suffer a slump, ANALYSIS OF THE WTI VERSUS THE while demand for consumable goods and services HISTORIC RETIREMENT AGE remains level. The expected result would be price Table 2 gives calculated values for the WTI based inﬂation. on data for the years 1976 –98, yearly productiv- To the extent that the retirement decision is ity improvements for the same period, and me- dependent on the real value of assets accumu- dian retirement ages for Canada for the years lated versus the current cost of goods and ser- 1976 –95 (Statistics Canada 1999). Statistics vices, then it is clear that some baby boomers Canada has published data on age of retirement might be forced to postpone their exit from the only since 1976, so no earlier periods could be workforce (see also Schieber and Shoven 1994). analyzed. Employers, as well as governments, would also be A linear regression model of the average retire- expected to provide incentives for later retire- ment because there would be a decline in the supply of labor (Statistics Canada 1996, p. 39). Table 2 In other words, the baby boomers might be forced Median Retirement Age and Wealth Transfer to adjust to new ages of retirement that would continue to allow a constant wealth transfer from Index (Adjusted and Unadjusted) for a stable workforce to all dependent Canadians. 1976 –1998 Productivity WTI- Year MRA Increase WTI Adjusted Figure 1 1976 65.00 1.89% 2.962 2.962 1977 64.92 0.00 2.927 2.872 Distribution of Historical and Projected 1978 65.00 0.00 2.847 2.793 Population by Age Group 1979 64.92 0.46 2.747 2.696 1980 64.83 0.00 2.684 2.622 1981 64.92 0.92 2.617 2.556 1982 64.83 2.28 2.721 2.633 1983 64.67 3.13 2.708 2.559 1984 64.75 1.73 2.639 2.413 1985 64.67 0.43 2.571 2.307 1986 64.58 0.43 2.515 2.267 1987 63.25 0.00 2.488 2.232 1988 63.83 0.00 2.443 2.192 1989 63.33 0.43 2.441 2.190 1990 62.92 0.85 2.475 2.210 1991 62.58 1.68 2.582 2.284 1992 62.33 0.41 2.636 2.289 1993 61.83 1.65 2.632 2.274 1994 62.33 1.21 2.605 2.209 1995 61.75 0.40 2.583 2.159 1996 2.81 2.579 2.166 1997 0.78 2.551 2.073 1998 2.506 2.017 Source: Ofﬁce of the Superintendent of Financial Institutions (1998). Source: Statistics Canada (1999), Brown and Bilodeau (1999). 4 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 5, NUMBER 2 ment age was ﬁtted against the WTI, adjusted for Figure 3 annual labor productivity improvements (see Fig- Model 2, 1982–1995 ure 2). The median retirement age for a particular Overlay Plots of Actual Versus Predicted year was regressed on the resultant WTI of that Average Retirement Age for Canada same year. Signiﬁcant overall regression was ob- with Lagged WTI tained with an R2 statistic of 0.55. The regression equation is Median Retirement Age (years) 55.40 3.47 Adjusted Wealth Transfer Index. A second regression model—also accounting for labor productivity improvements in the calcu- lation of the WTI—was ﬁtted (see Figure 3). The median retirement age for a year was regressed on the WTI lagged six years; that is, the retirement age of year t was regressed on the WTI of year t 6. Results for this model were impressive, with an R2 statistic of 0.91. The obtained regression equa- tion is the tax rate). This could take years. Finally, after Median Retirement Age (years) 52.77 a response decision has been reached, time would 4.22 Adjusted Wealth Transfer Index (lagged). be required for implementation of the suggested course of action. For example, if the WTI de- Lagging the WTI used in the regression model is clined, it might be possible for governments to plausible because individuals, employers, and lower taxes, for manufacturers to lower prices, or governments all need time to make adjustments for employers to enhance pension beneﬁts. Any to accommodate new realities. The lags involved of these actions would allow earlier retirement. can basically be placed into three categories: rec- As a second example, with the impending de- ognition, decision, and implementation lags. mographic proﬁle where the baby bust generation First, the agent involved (individual, employer, or will be the source of labor, it might be expected government) needs to identify and recognize that that both employers and governments would offer a wealth transfer shift has occurred (either up- late retirement incentives. However, they would ward or downward). Once this has been identi- require time to identify the need and then to ﬁed, time is needed to respond (e.g., by changing implement the incentives. It would also take time for the employee to factor these incentives into his or her retirement decision. Thus, a six-year Figure 2 time lag is completely plausible. Model 1, 1976 –1995 Finally, a linear regression model that did not Overlay Plots of Actual Versus Predicted account for labor productivity improvements was Average Retirement Age for Canada ﬁtted using lagged values of the WTI observed during the period 1976 –95 (see Figure 4). The results show that the regression is still signiﬁcant (R2 0.84) and that the WTI is a useful predictor of the median retirement age (although not as good as model 2). The regression equation is Median Retirement Age (years) 46.77 6.24 Wealth Transfer Index (lagged 6 years). The regression results for the three models, particularly the lagged WTI model, show the ex- A MACRO-ECONOMIC INDICATOR OF AGE AT RETIREMENT 5 Figure 4 (1984, 1989, 1995b) up to 1994. We then use an Model 3, 1976 –1995 ARIMA1 time series methodology to project these Overlay Plots of Actual Versus Predicted rates to 2006, after which the rates are held con- Average Retirement Age for Canada stant. The participation rates are segregated be- Ignoring Productivity tween the sexes and different age classes while the unemployment rate is obtained for the entire adult population. By knowing the number of peo- ple in the various age and sex categories, the model forecasts the number of employed and un- employed Canadians to 2041. The model assumes an annual increase in pro- ductivity consistent with the historical increase from 1976 to 1998 (in terms of 1986 dollars). The productivity increase during this period averaged 0.9% compounded per annum (pa). The WTI (ad- justed) to 2041 is then found using the projected population and employment data, with the labor force component adjusted to reﬂect productivity improvements. istence of a strong positive correlation between Using the six-year lagged regression we ob- the WTI and median retirement age. The WTI tained in the previous section (model 2), we are quantiﬁes the economic force that “decides” the able to project the median retirement age in Can- average age at retirement as a ratio of consump- ada to 2047. The result is displayed in Figure 5. If tion demand to production supply. Because of the retirement age rises, it is assumed that these this deﬁnition of the WTI and the regression re- workers have the labor force participation rates of sults, it might be plausible to infer a causal rela- those ages 60 – 64. tion between the WTI and median retirement age. From Figure 5, we can see that the median retirement age is projected to decrease generally until 2017, when it reaches a local minimum of PROJECTIONS ON FUTURE RETIREMENT AGE IN CANADA The previous section provides a model (model 2) 1 For more information on the ARIMA (autoregressive integrated mov- with which to project the retirement age in the ing average) process, see chapter 4 of Time Series and Analysis: Forecasting and Control (Box and Jenkins 1976). future. To this end we also need to project the WTI into the future. Statistics Canada (1994) has projected the 1993 Canadian population to Figure 5 2041 under four different sets of assumptions Median Retirement Age in Canada (low-growth, medium-growth, and two high- growth projections). This paper employs the me- with Productivity Improvement of 0.9% pa dium-growth projection (projection 2); it is con- (1996 –2047) sidered to be the most realistic, and Brown and Bilodeau (1999) used the same assumption in their paper. These data provide us with informa- tion on the number of the young (ages 0 –19), adults (ages 20 – 64), and elderly (ages 65 and older) in Canada to 2041. To determine the number of employed and un- employed adults, we use a method similar to Brown and Bilodeau. Historical participation rates and unemployment rates for various age and sex groups are available from Statistics Canada 6 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 5, NUMBER 2 60.3 years. After this date, the increase in the Figure 7 number of elderly and the decrease in employed Median Retirement Age in Canada with adults results in a higher median retirement age Productivity Improvements of 0.0% pa as workers must stay longer in the workforce to (1996 –2047) achieve a constant WTI. The increase is projected to last until 2034, when the median retirement age is 60.9 years. After that, the retirement age is again projected to decrease. In 2041, the median retirement age is forecast to be 60.6 years; it will be 60.0 years in 2047. We also show future projected retirement ages with 1.5% pa productivity growth (Figure 6) and no productivity growth (Figure 7). Finally, in Fig- ure 8, we show that the annual rate of productiv- ity growth required for no increase in retirement age is 1.29% pa. This projection of the expected retirement SUMMARY age is consistent with a philosophical view that There appears to be a very strong positive corre- government transfers, including education, un- lation between the WTI and retirement age based employment insurance, and social security, on Canadian historical data for the years 1976 – represent wealth transfer. But wealth can be 95. It is plausible that because of the time re- transferred only after it has been created, and quired for recognition, decision, and implemen- then only in the amount that has been created. tation, the effects quantiﬁed by the WTI will be This paper concludes that, historically, workers lagged several years (apparently close to six have retired at the earliest possible age that was years). affordable given the limits on the potential trans- The regression model obtained (model 2) was fer of wealth. The paper further concludes that used to project future retirement ages. It indi- this will continue to be true, whether legislated by cates that the median retirement age in Canada government or not. If true, the retirement age will generally decrease from 61.75 in 1995 to 60.3 experienced by the workforce is just another re- years in 2017, increase slightly to a local maxi- sultant variable in a macro-economy that must mum of 60.9 years in 2034, and then decrease operate in balance: that is, the variable “retire- once again to 60.0 years at the end of the projec- tion period, 2047. Figure 8 Median Retirement Age in Canada with Figure 6 Productivity Improvements of 1.2895% pa Median Retirement Age in Canada with (1996 –2047) Productivity Improvements of 1.5% pa (1996 –2047) A MACRO-ECONOMIC INDICATOR OF AGE AT RETIREMENT 7 ment age” is just another balance-point variable ———. 1999. Historical Median Retirement Ages in Canada. that will be decided by economic realities, not Ottawa: Ministry of Industry, Science and Technology. TASK FORCE ON RETIREMENT INCOME POLICY. 1979. “The Retirement government legislation. Income System in Canada: Problems and Alternative Pol- icies for Reform.” Ottawa: Ministry of Supply and Services. REFERENCES DISCUSSION BOX, GEORGE, E. P. AND G. M. JENKINS. 1976. Time Series and Analysis: Forecasting and Control. San Francisco: Holden- Day. BERNARD DUSSAULT* BROWN, ROBERT L. 1999. Economic Security for an Aging Cana- This paper has much more profound implications dian Population. Society of Actuaries Monograph M-RS99-2. Schaumburg, IL.: Society of Actuaries. than its title suggests. The authors are to be con- BROWN, ROBERT L., AND C. BILODEAU. 1999. “The Canadian Wealth gratulated for being so proliﬁc in the production Transfer Index.” Proceedings of the Canadian Institute of of quality research papers that beneﬁt not only Actuaries, 1996 –97 28(2): 435–54. the actuarial community but also all the players FOOT, D. K. 1982. Canada’s Population Outlook. Canadian In- involved in the evolution toward better designed stitute for Economic Policy. Toronto: James Lorimer. social security schemes around the world. MCDONALD, L., AND E. B. CARTY. 1980. “Effect of Projected Popu- I naturally tend to consider that the statutory lation Change on Expenditures of Government.” In The Retirement Income System in Canada: Problems and Al- actuarial reports on social pension schemes (such ternative Policies for Reform, vol. 2, 16-1–34, ed. Task as OASDI and SSI in the United States, and CPP, Force on Retirement Income Policy. Ottawa: Minister of OAS, and GIS in Canada) are the end of the road Supply and Services. in terms of information required for government OFFICE OF THE SUPERINTENDENT OF FINANCIAL INSTITUTIONS. 1998. Can- decisions on the ﬁnancing of such schemes. I, ada Pension Plan Seventeenth Actuarial Report as at 31 therefore, tend to think that all other reports and December 1997. Ottawa: Ministry of Finance. papers relating to the analysis of those schemes SCHIEBER, S. J. AND J. B. SHOVEN. 1994. “The Consequences of are the useful policy tools required in examining Population Aging on Private Pension Fund Saving and Asset Markets.” NBER Working Paper No. 4665. Cambridge, MA: various options for the design of the beneﬁts pro- National Bureau of Economic Research. vided through, and the underlying ﬁnancing of, STATISTICS CANADA. 1984. Labour Force Annual Averages, 1975– such schemes. My discussion is, accordingly, in 1983. Catalogue No. 71-529. Ottawa: Ministry of Industry, two parts relating to ﬁnancing and social policy, Science and Technology. respectively. ———. 1986. “Longevity and Historical Life Tables (1921–1981) (Abridged).” In Canada and the Provinces. Financing Ottawa: Ministry of Supply and Services. ———. 1989. Labour Force Annual Averages, 1981–1988. Cat- The paper is not just another one providing good alogue No. 71-529. Ottawa: Ministry of Industry, Science food for policy thought. Before I start comment- and Technology. ing on the main target of the paper, namely re- ———. 1994. Population Projections for Canada, Provinces tirement age, I want to bring the following extract and Territories, 1993–2016. Catalogue No. 91-520. Ot- to the attention of chief actuaries responsible for tawa: Ministry of Industry, Science and Technology. the long-term ﬁnancial valuation of social secu- ———. 1995a. Life Tables, Canada and the Provinces, 1990 – 1992. Catalogue No. 84-537. Ottawa: Ministry of Industry, rity schemes: Science and Technology. ———. 1995b. Labour Force Annual Averages, 1989 –1994. Furthermore, because the much smaller baby Catalogue No. 71-529. Ottawa: Ministry of Industry, Sci- bust generation is now the only source of labor, ence and Technology. production in the economy could suffer a slump, ———. 1996. Canada’s Changing Retirement Patterns: Find- while demand for consumption goods and services ings from the General Social Survey. Catalogue No. 89- remains level. The expected result would be price 546-XPE. Ottawa: Ministry of Industry, Science and Tech- inﬂation. nology. ———. 1997. Report on the Demographic Situation in Canada, 1996. Current Demographic Analysis. Catalogue No. 91- 209-XPE. Ottawa: Ministry of Industry, Science and Tech- *Bernard Dussault is a Consulting Actuary, 1797 Rue des Arbres, nology. Orlean, ON KIE 2T7, Canada, e-mail: firstname.lastname@example.org. 8 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 5, NUMBER 2 Not only do actuarial reports normally provide turn, should slow the pace of declining mortality. a relatively low volume of documentation on the These are examples of what an assumption’s rationale used by the valuation actuary for the framework should cover. Productivity (i.e., real determination of key economic and demographic GDP per capita or real salary increase) may ap- assumptions, but those assumptions also appear pear to be the only isolated assumption not to be chosen without reference to an explicitly strongly correlating with any other key economic and substantially documented framework of or demographic assumptions. However, as aging benchmark values. No surprise then that Brown, increases the average age of the labor force, one Damm, and Sharara are the ﬁrst (to my knowl- should strongly consider that it might also reduce edge) to alert those valuation actuaries implicitly productivity. that the rationale for their inﬂation assumption As demonstrated in the paper, a sustained an- might be somewhat ﬂawed. nual productivity rate close to 1% would more All assumptions are probabilities to which a than offset the effect of aging on the balance certain degree of uncertainty applies. This degree between demand and supply. Without such pro- is normally relatively high, with at least one ex- ductivity, Brown, Damm, and Sharara say “some ception: fertility. This also applies to the pro- baby boomers might be forced to postpone their jected worldwide aging of populations. Indeed, exit from the labor force.” However, they would the Canadian population aging projected for 2030 not be naturally forced, willing, or in a position results mainly (about 90%, the remaining 10% (considering the above suggested correlation be- relating to declining mortality) from the world- tween aging and productivity) to increase their wide decrease in fertility over the past 30 years productivity in order to reset the demand/supply and the projected sustained low fertility for the imbalance or to bring inﬂation down. 21st century. It is precisely the population aging, which is projected with a relatively high degree of cer- Social Policy tainty, that will inevitably induce an imbalance Actuaries in North America started over 20 years between the demand for and the supply of goods ago to warn governments and the public that the and services. In the Old Age Security (OAS) Pro- population is aging and projected to continue so gram, Fourth Actuarial Report (OSFI 1999), the at a steeper rate during the 21st century, and that age dependency ratio (ADR), at a level of about this would increase social security costs. From 20% for year 2000, is projected to climb gradually the very beginning, this sparked discussions on to 35% by 2030 and to 44% by 2100. Those are increasing the normal retirement age (NRA) as a tremendous increases. My personal reaction to means to contain such costs. In the U.S., an in- the authors’ implicit warning is twofold: crease from 65 to 67 in the NRA was legislated under the OASDI in 1983 but to apply on a grad- ● The risk of inﬂation (i.e., higher than 3% on ual basis (over 24 years) starting only in 2003. average) might not have been properly ac- Although such a solution has also been dis- counted for in the cash ﬂow projections pre- cussed at large in Canada, the NRA has so far sented in the Fourth OAS and 17th CPP statu- remained 65 for Canadian social security tory actuarial reports. schemes. Various other means have been exam- ● Policymakers should urge governments to pre- ined to deal with the projected increases in CPP scribe a ceiling (e.g., annual salary increase) on costs. One must note that OAS and GIS projected the price-related indexation generally applying costs decrease in the long-term as these are price- to pensions in pay. related rather than earnings-related programs. In Understandably, a given assumption (e.g., inﬂa- this connection, one may refer to Graph II.2 of tion) should not be reviewed and revised in iso- the OAS Fourth Actuarial Report at www.osﬁ- lation. The reduced annual demographic growth bsif.gc.ca (Ofﬁce of the Chief Acturary, Actuarial resulting from lower fertility should normally en- Reports). The CPP was reformed in 1997 with the tail lower economic growth (consistent with a understanding that an increase in the NRA had presumed natural equilibrium between demand been considered but not retained at this time as a for and supply of goods and services), which, in means to reduce future CPP contribution rates. A MACRO-ECONOMIC INDICATOR OF AGE AT RETIREMENT 9 Nonetheless, the macro-economic indicator of amining the pattern of the projected MEIAR val- age at retirement (MEIAR) conceived by the au- ues. On the basis of a starting value close to age thors is not, and does not pretend to be, an indi- 62 in 1996, the MRAC projected using the MEIAR: cator to be used to index the NRA under the Canadian social security schemes (OAS, GIS, ● decreases until 2016 (to close to age 60), CPP). One must ﬁrst keep in mind that this indi- ● increases to close to age 61 by 2036, and cator is in respect of the median retirement age in ● drops back close to age 60 by 2046. Canada, that is, resulting from all employed per- Many people (even experts) actually link the sons in Canada as opposed to the NRA under the 2016 –2041 hump to the baby boomers. That is OAS, the GIS and the CPP, which is simply set at erroneous. MEIAR is a compounded demograph- age 65 presently. In other words, one should re- ic-economic indicator. Baby boomers are the re- frain from falsely concluding that the median re- sult of an exclusively demographic, not economic, tirement age in Canada, projected using the ﬂuctuation. One must accordingly refer to a MEIAR, would stabilize the pension costs under strictly demographic indicator to assess the im- OAS, GIS and CPP. In fact, it would not for any of pact, be it demographic or economic, of the baby these three programs. However, the MEIAR might boomers. The ADR values shown in Table actually be a consistent tool for indexing the NRA VII.B.10 of the Fourth Actuarial Report on OAS under the OAS and the GIS if the objective of are ideal in that respect. As mentioned earlier in these programs were to transfer income between this discussion, ADRs shown in Table VII increase active workers and seniors at a relative level con- without interruption from year 2000 to 2100. sistent with the stabilization of the demand/sup- That table shows values for a limited number of ply ratio. years only, but the complete year-by-year demo- However, the pattern of the projected median graphic projection (of the Fourth OAS Report) retirement age in Canada (MRAC), as shown in supports this statement. Figure 2 of the paper, deserves some attention Still, the most striking conclusion that can be and explanation. If one puts the projected MRAC drawn from the projected MRAC values is that, (Figure 2) and the projected OAS costs (Graph assuming a productivity rate close to 1%, not only II.2 of the Fourth Actuarial Report on OAS) side would the current demand/supply ratio remain in by side, the similarity of the two curves is strik- balance, but also the actual median retirement ing. In fact, it is doubly striking as they have the age would be reduced from 62 to 60. However, it same shape with a six-year lag. This is not a must be pointed out that with a nil productivity coincidence even if Figure 2 represents an age rate, the demand/supply ratio would be stabilized and Graph II.2 a paygo rate. The reason for the only provided an increase of over ﬁve years in the similarity is that a paygo rate is an ADR dis- median retirement age, as reported in Brown and counted for productivity, whereas the Wealth Bilodeau (1997). Transfer Index (WTI— used as a template to de- The reason for the huge impact of a mere 1% velop the MEIAR and the MRAC)—is an ADR productivity rate on the retirement age is that adjusted to include the youth dependency rate as productivity applies to the entire national pay- well as weights for relative (youth, juniors, and roll, whereas the portion of the payroll earned seniors) dependency costs. by workers over age 60 is quite small. This is The inclusion of youth in the WTI could disturb because the population in the 60 –70 age range materially the pattern of the MRAC curve, but it represents less than 20% of the population in does not because of the small and reduced the 20 –70 age range (due to mortality increas- weights of children in the WTI, both in terms of ing by age) and because employment rates are their proportion and of their dependency costs much lower for the 60 –70 age range than for compared to seniors. Such similarity does not the 20 – 60 age range. exist between the MRAC and the CPP projected paygo costs because the CPP is an earnings-re- lated (standard of living) rather than a price- Conclusions related (purchasing power) program. No one knows what the productivity rate will be Another point deserves some attention in ex- on average in the long term. However, productiv- 10 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 5, NUMBER 2 ity may reasonably be expected to be lower in the will just result in reduced standards of living. This long run in light of the aging of the population. will likely be the economic outcome, and its effect In any event, in a context of low or nil produc- on social security programs can only be cost in- tivity, where the objective is to stabilize the de- creases and/or beneﬁt reductions. mand/supply ratio, the average median retire- ment age would need to be increased whenever References the WTI increased. In this case, it is not guaran- BROWN, R. L. AND C. BILODEAU. 1997. “The Candadian Weath Transfer Index,” in Proceedings of the Canadian Institutes teed that the labor market could absorb such an of Actuaries 28(2): 435–54. increase. The multiphased transition approach OFFICE OF THE SUPERINTENDENT OF FINANCIAL INSTITUTIONS. 1999.“Old for retirement, discussed inter alia by Robert L. Age Security Program, Fourth Actuarial Report.” Ottawa: Brown in one of his several research papers (In- Ministry of Finance. Online at http: //www.osﬁ-bsif.gc.ca. stitute of Insurance and Pension Research, Uni- Additional discussions on this paper can be submit- versity of Waterloo, 98-05), is likely the only tan- ted until October 1, 2001. The author reserves the gible solution for dealing optimally with unknown right to reply to any discussion. Please see the Sub- but actual future variations in productivity and mission Guidelines for Authors on the inside back employment rates. And any eventual failure of cover for instructions on the submission of discus- the labor market to absorb an increase in demand sions.