Presentation to the London Actuaries Club: Mortality Projections for Canadian Social Security Programs

Good afternoon everyone and thanks to the London Actuaries Club for giving my colleague Shayne and I the opportunity to speak to you today. The purpose of the presentation today is to provide information on the methodology that our office employs to determine the best-estimate mortality assumptions for the actuarial reports that our office prepares. We delivered this presentation in the format of a webinar for the Canadian Institute of Actuaries in October. If you listened in on that webcast, I am sorry for the overlap. With the exception of a few updates on data with regards to COVID-19 at the end of the presentation, the content is identical. I think that everyone will agree that the subject of the presentation is fairly dry, especially given the fact that it is from 4 to 5 on a Thursday. Despite this, I hope that you can all appreciate that mortality is a very important aspect of actuarial work in many fields, including retirement, insurance and social security.

Office of the Chief Actuary (Slide 2)

I will start by providing you with a brief overview of our office, the Office of the Chief Actuary. The Office of the Chief Actuary, or OCA as we are often referred to, is a small group of about 40 individuals, most of whom are actuaries or have an actuarial background. We are an independent unit within the Office of the Superintendent of Financial Institutions. Although we are part of OSFI, which I am sure you all know, our mandate is distinct from the rest of OSFI. Our mandate is to conduct statutory actuarial valuations for a wide range of social security programs and federal public sector pension and insurance plans. The programs that fall under our responsibility are listed on this slide. All our reports are tabled in Parliament by an appropriate minister.

As a result of our statutory mandate, our office and more specifically our Chief Actuary who is currently Assia Billig, are solely responsible for the assumptions, content and actuarial opinions of our reports. Further, the legislations require us to use best-estimate assumptions in all our reports with no margins for adverse deviation.

Since we work with large social security programs and pension plans, we have access to a wide range of data. As a result, we are in a position to develop mortality projections for the Canada Pension Plan, the Old Age security program as well as large federal public sector pension plans.

Over the course of this presentation, my colleague Shayne and I will provide you with information on how with develop our mortality projections and what factors we take into consideration.

Triennial Actuarial Valuation of the CPP (Slide 3)

Our office prepares statutory actuarial valuation reports on the Canada Pension Plan every three years. Under normal circumstances, the mortality assumptions prepared in the context of the triennial CPP actuarial report are used as a stepping stone for the other statutory reports that our office prepares.

For example, the population projections in the OAS statutory report, which is on the same triennial cycle as the CPP, are the same as the ones used in the CPP report. Also, mortality improvement rates used in the statutory report for most of the federal public sector pension plans are based on the assumptions from the most recently tabled CPP report.

However, it is safe to say that we are currently not in normal circumstances. Although the most recent CPP report was prepared and tabled in Parliament before the onset of the COVID-19 pandemic in Canada, it would be difficult to sit here today and explain our mortality assumption process without addressing the elephant in the room.

Although our office has started looking into potential impacts of the pandemic on mortality in the context of the next CPP report, we still have not conducted enough work to provide you with concrete views on the potential short-term and long-term effects. That being said, the last portion of this presentation will cover mortality considerations in the context of the covid-19 pandemic.

Triennial Actuarial Valuation of the CPP (Slide 4)

The most recent triennial report on the CPP was as at December 31, 2018, and it was tabled in Parliament in December 2019. As mentioned on the previous slide, this was before the onset of the COVID-19 pandemic in Canada.

To give you a bit of context on the work that we do for the CPP, our CPP actuarial reports provide, among other things, open-group projections of future inflows and outflows of the CPP over a projection period of over 75 years.

Our first step is therefore to project the general population by age and gender over the projection period. The mortality assumption of the general population is an important component of our population projections. Separate mortality assumptions are also developed for different subsets of the CPP beneficiaries populations to reflect mortality differences of these sub-populations with the general population.

These differences in mortality are accounted for through different starting mortality rates. The same mortality improvement rates are applied to all subgroups.

Before passing it on to Shayne, I just wanted to highlight that although we develop our mortality assumption for all ages groups, the main focus is for ages 65 and over since these are the ages that really drive the CPP costs when it comes to mortality.

I will now pass it over to Shayne who will provide you with additional details on our methodology.

Mortality Rates Projections for CPP (Slide 5)

Thanks Christine. I’m going to go over the general outline of how we develop our mortality rate projections for the CPP. First of all, we start with the population mortality from Statistics Canada. At the time of the valuation, the most recent available data was 2015. For retirees and survivors, we apply adjustment factors to that starting population mortality, while for disabled beneficiaries, we use the experience data. More information on these factors is provided in our actuarial study number 16 and Christine will also touch on them later.

The next step is the core of our assumption, which is the mortality improvement rates. We project these using a combination of backward and forward looking approaches. What that means is that we analyse historical improvement rates as far back as we can and consider both past and future potential impacts on mortality improvements, which we will talk about later on in the presentation.

Once we analyse this, we use a judgement-based approach to set the select and ultimate improvement rates by age and sex. For CPP30, the historical data we used came from the Canadian Human Mortality Database (CHMD) and STATCAN mortality rates for the Canadian population. The experience data on mortality rates by age from CHMD were available from 1921 up to 2011. The experience data on mortality rates by age from STATCAN were available up to 2015.

Once we have our assumption, we then conduct sensitivity analysis to see the impact on the CPP minimum contribution rates.

In the end, we don’t project life expectancies, what we project are the mortality improvement rates and the life expectancies are simply a result of these projections.

Mortality Rates are Higher than Expected in CPP27 (Slide 6)

Let’s now take a look at our starting point, which is the 2015 STATCAN mortality rates by age and sex. This slide shows the life expectancies in 2015 compared to what we projected them to be in the previous report, CPP27. Historically, we had consistently been over projecting mortality, which means that our valuation of the mortality improvement assumptions were always lower than the actual experience. This meant that actual life expectancies were higher than we projected them to be. However, for the first time in decades, our starting mortality rates for the CPP30 are higher than what was projected under the previous CPP report. As a result, the life expectancy for calendar year 2015 is slightly lower than expected.

We can see this on the table. Life expectancy at age 65 is 0.4 years lower for males and 0.2 years lower for females. At age 80, the decrease is 0.1 years. Let’s now take a look at why that is, what has been happening to life expectancy historically and whether Canada is the only country experiencing this slow down?

Life Expectancy at Age 65 – Period Life Expectancy at Age 65 for Males (Slide 7)

The next few slides will take a look backwards and analyze the historical experience as well as some of the past drivers of mortality. An important step in determining our mortality assumption is analysing and understanding past trends in mortality. This can be done by looking at the evolution of life expectancy over time. During our assumption-setting process, we also compare where Canada is in relation to other countries. Also, even though we develop mortality rates for all ages, we will focus mainly on the older age groups of 65+, which, as Christine mentioned, represent CPP’s main cost.

The graph shows the historical calendar life expectancy for males in Canada compared to 5 other countries with a population of more than 8 million and some with the highest life expectancy in the world such as US, Japan, France, Finland and the UK. As we can see, Canada (in red) fares quite well internationally and is among the top 3 countries listed.

In the 50s, US and Canada had higher life expectancy than Europe and Japan. From the mid-70s, other countries started to catch up. Over the last decade, these three countries have very similar life expectancy for males and in 2015 Canada is among the highest.

Another phenomena well seen on this slide, is that life expectancy is no longer increasing at the same rate as before, which is well pronounced for UK and US (in the green and brown). Canada, as we will see later, experiences the same phenomena but not to the same extent.

In 2015, life expectancy at age 65 for Canadian males was 19.3, only lower than Japan with 19.4 and France following closely behind at 19.1.

Life Expectancy at Age 65 – Period Life Expectancy at Age 65 for Females (Slide 8)

We have the same trend for females. From the 1950s, there was a continuing increase in life expectancy at age 65. However, the rate of increase is starting to taper off starting in 2010.

It is interesting to note that males in Canada, France and Japan have very similar life expectancies, while females in both France and Japan are significantly ahead of Canada (23 and 24.3 compared to 22.1). For Finland, male life expectancy is lower than in Canada, but for women, the numbers are very close.

Something important to think about when comparing different country life expectancies is that there will always be specific characteristics that will result in differences, such as: infant mortality rates, diet, income inequality and lifestyle. We also need to keep in mind that the life expectancy masks a lot of differences by age, health, etc. which we will look at on the next slide.

After age 85, Canada along with Japan and France have the lowest mortality rates (Slide 9)

Life expectancy is an aggregate number which masks variations by age. Our assumptions are developed on the level of mortality rates by age and sex, and life expectancies are a result. In order to determine how mortality rates may evolve in the future, it is interesting to see for which age groups Canada has room to improve compared to other countries.

This slide shows mortality rates by age group for ages 65 and over for 11 countries with high longevity and a population of more than 8 million. We’ve added a few more countries for comparison, such as Germany, Sweden, Switzerland, Australia and Spain. Canada is in the red. Countries with bars that are below the red horizontal line have mortality rates that are lower than Canada while countries with bars that are above the red horizontal line have mortality rates that are higher than Canada.

As you can see, while there is not much difference in mortality rates among countries for age groups from 65 to 84, Canada (red bars) does pretty well, along with France and Japan but has some room to improve.

As we move to age groups 85 to 94, Canada becomes one of the top three countries with the lowest mortality rates along with Japan (light blue bars) and France (dark blue bars).

Cancer and Heart Diseases are the Leading Causes of Death in Most Countries (Slide 10)

So far, we’ve looked at life expectancy through time and have observed a slow down and at mortality rates by age and have observed that although Canada has some room to improve, it is doing well internationally. We will now analyze the main drivers for these trends by looking at causes of death.

This slide presents mortality rates for four of the top causes of death: cancer, heart diseases, cerebrovascular diseases and influenza and pneumonia for the age group 75-84 (both sexes combined). Red bars are rates for Canada, black bars are for the country with the highest rates among those shown on the previous slide, and green bars are for the country with the lowest rates.

Looking at cancer, UK has the highest mortality rates. However, Canada doesn’t fare very well: it shares the second highest cancer mortality rates.

On the other hand, for heart diseases, the only two countries with rates lower than Canada are France and Japan. And for stroke, it is only France.

Influenza and pneumonia mortality rates provide an illustration that even in Japan people are not immortal. Although Japanese seniors may die less from cancer and heart diseases, there are much more likely to die from pneumonia than seniors in other countries. While we have room to improve in cancer, our heart disease and stroke mortality rates compare well internationally.

Improvements in mortality related to heart diseases have been significant over the last decades (Slide 11)

Now, let’s look at causes of death strictly for Canada and see if we can explain the slowing down of the pace in life expectancy increases. This slide shows mortality rates for ages 65 + for the top causes of death.

Right away we can see that heart disease mortality has improved significantly over the last few decades. This is a large contributing factor to the significant increases in Canadian life expectancies at age 65 that have been observed. These rates were improving at around 4% per year at ages 65 and over for males and females (as will be shown on the next slide).

However, as we can see, the pace of improvements for heart diseases for males recently decreased significantly. The black bar shows the slope of the lines changing since 2010.

We also see the continuing decreasing trend in deaths from stroke (yellow), cancer (blue), and lung cancer. The latter is most likely related to the reduction in smoking among males.

For women, mortality from cancer reduces slowly (Slide 12)

The picture for females is very similar to the one for males for heart diseases and stroke. Again, as you can see, heart disease mortality, while still going down, has started to decrease at a slower pace.

Unfortunately, mortality rates from cancer seem to be quite unchanged.

Further, deaths from lung cancer are on the rise. It is probably the result of cohorts of women who started to smoke later than men.

Historical Mortality Improvement Rates Long-Term Average (Slide 13)

Now that we have looked at causes of death and life expectancy and seen where the improvements have been slowing down, we move on to mortality improvement rates which are the crux of our assumptions for CPP. Looking purely backward, we can see that over the last 90 years (let’s take the longest available period since our projections are very long-term), the average mortality improvement rates for Canada were 1.0% per year for ages 65 and over and 0.7% per year for ages 85 and over.

Similar to many other countries, Canada saw declining rates of mortality improvements. The slowing down trend is also evident when looking at improvement rates. As we can see over the last 15 years (ending in 2015), the MIR for 65+ were 1.9% which has slowed down to 1.3% when just looking at the last 5 years. You can also see that improvement rates for the age group 65-74 in the last 5 years of 0.9% are even lower compared to 1.3% when using a 90-year average.

Slowdown in mortality improvements in recent years: a blip or a new trend? (Slide 14)

The same trend can be seen when looking at Old Age Security Program experience which covers more than 97% of the Canadian population over 65. This slide presents historical mortality improvements by age groups for ages 65 and over for three different periods of five years (2004 to 2009, 2009 to 2014 and 2014 to 2019). Males are in blue and females are in orange.

Here you can see the decreasing trend is especially pronounced for male OAS beneficiaries in the 65-69 age group: average MIRs were close to 2.0% from 2004 to 2014, and reduced to 0.4% from 2014 to 2019. Similarly, the average annual MIRs for female OAS beneficiaries age 65-69 were about 1.7 % from 2004 to 2014, and decreased to 0.4% from 2014 to 2019.

There is a decreasing trend in the annual mortality improvement rates for other age groups over age 65 as well. More information is available in the OAS Mortality Fact Sheet, which our office published in November 2020.

So generally speaking, we still observe improvements at ages 65 and over but at a decreasing pace compared to recent past. Now that we have covered our backward looking analysis and past drivers, Christine will talk about the forward looking analysis and future drivers we take into consideration when developing our assumptions.

Future drivers of mortality are not easy to quantify (Slide 15)

Shayne presented some of the data and indicators that we analyse for understanding the past. However, our methodology is not based on projections by cause of death, nor is it based on fitting and extrapolating past mortality data based on mathematical models such as the Lee Carter model. In our view, past trends are not necessarily reflective of what will happen in the future given that the drivers of mortality in the future will differ from those of the past. Rather than extrapolating past trends, we use a judgement based approach in developing mortality projections. We consider the historical trends and make an assessment whether these trends are likely to continue into the future, and we look at emerging factors and potential future drivers of mortality.

For example, in terms of assessing previous trends, as my colleague Shayne has discussed on previous slides, it seems that some of the “easy gains” due to reduction in mortality from heart diseases and reductions in smoking have been achieved and will be difficult to replicate. Success in cancer treatments, which is one of the main causes of deaths at older ages, have proven so far to be quite elusive.

When we turn our attention to future drivers of mortality, there is a lot of uncertainty. As you can see from the sample of drivers listed on the slide, future drivers of mortality could create both favorable and unfavorable experience. In addition to having both favorable and unfavorable factors, there is a lot of uncertainty surrounding these drivers and how they will interact with each other in the future. One interesting point from the slide as you can see is that we list pandemics as being a potential driver of future mortality. What was considered as being quite a theoretical potential driver of mortality has quickly become a reality. This highlights the uncertainty surrounding future mortality and how quickly things can change.

In addition to future drivers of mortality that may impact several countries, each country has economic and cultural characteristics that will influence future mortality trends. For example, given the difference in diet, climate, and lifestyle, is it realistic to expect that Canadian women will ever reach the level of mortality of Japanese women?

Over the next few slides, I will discuss some Canadians specific trends that may impact future mortality and which we consider in setting our assumptions.

By 2050, cost of obesity in Canada is projected to be 3 years of unrealized gains in life expectancy (Slide 16)

The first factor that I will discuss is future excess mortality resulting from obesity. Unfortunately, looking at the OECD countries, Canada ranks 29th out of 34 with almost 26% prevalence of obesity in adults. Obese individuals are at an increased risk of developing certain chronic conditions including hypertension, type 2 diabetes, cardiovascular diseases, some cancers and even premature death.

As shown on the slide, the OECD estimates that by 2050, Canadians are projected to live 3 years less compared to a scenario without obesity.

In 2017, in Canada, death rates due to opioid overdose were 1.6 – 2.1 times higher than in 2015 (Slide 17)

A recent sad development that we observe in Canada (following the similar but even more dramatic developments in US) is the increase in deaths from opioid overdose.

According to Statistics Canada, life expectancy at birth in Canada did not increase from 2016 to 2017 for either males or females, and it is largely attributable to deaths resulting from the opioid crisis. The chart on this slide shows the change in life expectancy in 2017 that is attributable to drug overdose. Based on this chart, you can see that drug overdoses led to a loss of life at birth of 0.1 years for men and 0.02 years for women.

Accidental drug poisoning deaths tend to occur among young adults, and therefore have a greater impact on life expectancy at birth. As shown on the chart, Alberta and BC were the largest contributors to this loss.

Even before the pandemic, the crisis was getting worse. According to the public health agency of Canada, rates of apparent opioid-related deaths at the national level remained high from July 2017 to March 2020.

Unfortunately, the COVID-19 crisis is exacerbating the opioid crisis in Canada. In September, Canada’s top doctor said that public health measures put in place to stop the spread of COVID-19 may have made it harder for people who use drugs to access needed support. In terms of data, several provinces have reported spikes in overdose deaths since the start of COVID-19. When we look at British Columbia for example, the numbers are staggering. Between March and November 2020, the number of overdose deaths in British Columbia have been higher than the number of COVID-19 related deaths, with the exception of November. The difference is also significant: between March and November 2020, there have been about 450 COVID related deaths compared to about 1,400 overdose deaths. Alberta is also a province where the opioid crisis is worsening as a result of the pandemic. The situation is not as critical as in British Columbia, but the number of overdose deaths has surpassed Covid-19 deaths for the first 10 months of 2020.

Mortality differences by socio-economic level diminish with age (Slide 18)

Another important factor that impacts mortality is inequality. Quite often inequality is measured by a difference in income. Individuals with higher income tend to have higher life expectancies than individuals with lower income, all else being equal.

For the CPP, we reflect inequality in our mortality projections by having mortality rates that vary by level of pension for CPP retirement beneficiaries. Our assumption is based on a historical analysis of the CPP retirement beneficiaries data.

The chart on this slide shows the life expectancies of CPP retirement beneficiaries by level of pension. As you can see on this slide there is definitely a difference in life expectancy among CPP retirement beneficiaries. At age 60 for example, males with higher pension are projected to live about 2.5 years longer than those with lower one. For females, this difference is slightly less than 2 years.

It also can be seen, that these differences diminish with age.

The gap in life expectancy by benefit level is stable over time (Slide 19)

In addition to reflecting inequalities in current mortality levels, trends in inequality matter as well. Over the past two decades, for both sexes, the difference in life expectancy at age 65 between CPP retirement beneficiaries receiving the maximum pensions and those with pensions less than 37.5% of the maximum has been stable.

As a result, in doing our projections we recognize differences in mortality by level of pension, but assume that the gap in mortality remains stable.

It is important to note that extreme situations such as the current COVID-19 pandemic can have an impact on the gap in life expectancy by socio-economic status. As we will see later in this presentation, based on preliminary data, individuals with lower socio-economic statuses seem to be more negatively affected by COVID-19. It is possible that this is a result of already existing differences in mortality by socio-economic status, or that the dynamics have changed and that inequalities have increased at least in the short term. It is therefore important to monitor this trend closely when extreme shocks occur that have the potential of affecting different subsets of the population differently.

Survivor Beneficiaries have a Much Higher Mortality than the Population (Slide 20)

CPP beneficiaries include survivors whose mortality is significantly higher than that of the general population. As can be seen on this graph, which shows the ratio of mortality rates between CPP survivors and the general population, mortality rates for survivors at age 65 for example are more than 30% higher than that of the general population.

One reason might be that survivors are deeply affected by the loss of their spouse, especially at the older ages where the survivor may already be in a weakened physical and emotional condition. Also, in some cases, one could assume that losing part of the primary source of income and social support adds stress for survivors.

For our projections, we adjust survivor mortality rates to reflect differences between survivor mortality and that of the general population based on historical experience.

Now that we have covered some of the factors that we consider in setting our mortality assumption, I will pass it over to Shayne who will provide information on our CPP30 final assumptions.

CPP30 Assumes that Mortality will Continue to Improve but at a Slower Pace – Historical and Projected Mortality Improvement Rates – Males (Slide 21)

As a result of the discussed considerations, the underlying assumption of our latest actuarial report is that mortality will continue to improve in the future but at a slower pace than in the past, with an ultimate MIR assumption of 0.8%.

This graph shows the historical and projected improvement rates as a 15-year average for males.

As can be seen on the graph, improvement rates have slowed down since 2011, which has been observed in many other countries. There is the same pattern for females but mortality improvement rates are at lower levels.

The mortality improvement rates for the last 15 years are gradually transitioned to the ultimate values in 20 years and held constant thereafter at 0.8%, which are the same as for CPP27.

CPP30 Assumes that Mortality will Continue to Improve but at a Slower Pace - Historical and Projected Mortality Improvement Rates – Females (Slide 22)

The same can be seen for females. It is assumed that the gap between male and female mortality rates will continue to diminish. At the same time, we believe, that male mortality rates will be continue to be higher than those of females, that is women will still live longer than men in the future.

CPP30 Annual Mortality Improvement Rates (Slide 23)

This shows the assumed mortality improvement rates starting point, transition period and ultimate for the different age groups. The ultimate rates are set at 0.8% in 2035 for all ages below 90, 0.5% for ages 90-94 and 0.2% for ages 95+.

So why 0.8%? Well to summarize the presentation so far, looking backward, the last 90 years of historical data were kept in mind for age 65+ (1.0%) and 85+ (0.7%). More recently, the experience data for both the Canadian population and OAS beneficiaries showed a downward trend in the pace of mortality improvement rates, especially in the last 5 years. We also observed a slowing down in improvements in causes of death, especially for heart diseases.

Looking forward, the future drivers of longevity were considered; however it is not possible to quantify the impact of each factor on future mortality improvement rates. The list of such factors could be long, encompassing areas such as medical advances, environmental changes, changes in lifestyle, emergence of new diseases, access to and quality of health care and long-term care, etc. Future improvements in mortality are unlikely to mirror the past due to the prior achievement of “easy gains”, as well as emerging causes of death being more difficult to address, such as diabetes and cancer, as seen in the earlier charts. In particular, it can be seen that a large part of past improvements came from the reduction of mortality from heart diseases. At the same time, there is a pronounced reduction in improvements over the last decade. On the other hand, a lot of non-negligible drivers are also putting downward pressure on future mortality improvement rates, such as increasing drug resistance, pandemics such as COVID19, natural and man-made disasters and increasing proportion of older people and the resulting pressure on quality and availability of health and long term care. Income inequality and lifestyle choices are other factors that are challenging to predict.

Consequently, based on what is known from the past and an analysis of potential future drivers, it was assumed that the ultimate mortality improvement rates would revert to a level of 0.8%, which is close to the long-term average.

Evolution of CPP Mortality Projections over 7 Actuarial Reports – Projected LE for those aged 65 – Male Canada (Slide 24)

As a result of our MIR assumptions, we can project cohort life expectancy. Shown in the chart are life expectancy for Canadian males aged 65 according to the assumptions used in our actuarial reports for the Canada Pension Plan. Each line represents the projection based on the mortality assumptions of the corresponding report.

As you can see, our mortality assumptions kept reflecting exceptional lower mortality experience than expected between each report. In 2012, we started to see the break in this trend.

In the 30th Report there is a small decrease in life expectancy in short term compared to the 27th report, but long term projections are the same.

In 2000, we projected that in 2019, the cohort life expectancy of males at age 65 would be 18.3 years. In 2018, this projection increased by 3.1 years to 21.4, and it is further projected to increase to 23.3 in 2050.

Evolution of CPP Mortality Projections over 7 Actuarial Reports– Projected LE for those aged 65 – Female Canada (Slide 25)

We observe a similar picture for women.

In 2000, we projected that in 2019, the cohort life expectancy of a female at age 65 would be 21.5 years. In 2018, this projection increased by 2.4 years to 23.9, and it is further projected to increase to 25.6 in 2050 (2.3 years higher than men).

Relative impact of different mortality assumptions on the base CPP minimum contribution rate (Slide 26)

We do spend a lot of time and effort developing mortality assumptions. But of course, these are assumptions and not predictions. Thus, it is important to quantify the impacts of alternative mortality assumptions on the cost of the social security plans. We do this by conducting sensitivity analysis.

As shown on the slide, if we double the assumed ultimate mortality rates, the cost of the base plan will go up by 3%. On the other hand, if we assume immediately no future mortality improvements, the cost of the plan will decrease by 7%. The last test indicates the cost of future mortality improvements to the CPP.

Now, that we’ve covered our final mortality assumptions for CPP30, I will pass it over to Christine who will talk to you about the impact of COVID-19 on mortality projections.

COVID-19 Leading Cause of Death in 2020 (Slide 27)

When we compare the number of deaths attributable to COVID-19 in 2020 with the leading causes of death in 2019, we can see that COVID-19 will constitute a leading cause of death for 2020 once the numbers on causes of death are compiled and published by Statistics Canada.

As we can see on the chart on this slide, there were 15,600 deaths due to COVID-19 in 2020. In 2019, the 3rd, 4th and 5th leading causes of death generated less deaths that this.

COVID-19 Impact – Excess Deaths in Canada (Slide 28)

Statistics Canada has been publishing weekly provisional data on the 2020 adjusted number of deaths in Canada and excess mortality.

The graph on this slide shows expected and actual number of deaths by week in 2020 based on Statistics Canada’s data and methodology. The orange line shows the expected number of deaths and the blue line shows the actual number of deaths. The two dotted lines show the lower and upper bound prediction of the expected number of deaths based on Statistics Canada’s methodology.

When looking at the graph, a few things stand out:

• First, the actual number of deaths, although within the prediction interval, were lower than expected before the onset of the pandemic. This is an indicator that without the COVID-19 pandemic, 2020 may have been a good year for mortality improvement rates.
• Second, and the most obvious one, there is a spike in deaths resulting from the first wave of the pandemic. During the 12-week period between March 31st and June 12th, there were approximately 8,600 excess deaths, with 50% of these deaths occurring in Quebec.
• Third, although within the prediction interval, the actual deaths following the first wave were higher than expected. During the 21-week period between June 13th and November 6th, there were about 4,900 excess deaths, with only 10% of these deaths occuring in Quebec. There was a significant increase in excess deaths in the Prairies and BC, which is most likely a combination of COVID-19 and drug overdoses resulting from the worsening opioid crisis.
• Although the data is incomplete to measure the full impact of the second wave, it has proven to be less deadly than the first one so far.
• Based on Statistics Canada’s excess mortality analysis, 2020 deaths up to October 31st are about 5% higher than expected. We will need more complete information by age and sex for the entire year to assess the impact on the 2020 life expectancy. That being said, a uniform increase by age of 5% in 2019 mortality rates corresponds to a decrease in life expectancy of about half a year for males and females.

Aggregate Numbers are Not Enough (Slide 29)

As mentioned previously, extreme situations such as the current COVID-19 pandemic can have an impact on the gap in life expectancy by socio-economic status. Tracking this gap and assessing its temporary or permanent nature is an important step of the mortality assumption process, especially in the context of social security programs that cover a wide and diverse population.

Aggregate numbers on COVID mortality measures such as the ones provided on the previous slides do not provide much insight on which subsets of the population are more at risk of contracting the virus or of developing complications or dying as a result of contracting the virus.

The data and research on this is still at the early stages. However, preliminary data is showing that certain groups are more at risk.

Certain at-risk groups were identified very early on such as older individuals as well as individuals with pre-existing health conditions. In Quebec, during the first wave of the pandemic, it also became clear early on that seniors living in long-term care homes in the province were considered very high risk.

As the pandemic is evolving and more data is becoming available, other at-risk groups are being identified. Based on this data, socio-economic status and ethnicity are considered to have an impact on COVID-19 mortality rates.

Higher Short-Term Mortality Rates (Slide 30)

One thing is certain at this point: Mortality rates in the short-term (meaning 2020 and most likely 2021) will be higher than expected and than in the recent past. Complete data is not available yet to easily measure the effect of the pandemic on the 2020 life expectancy. That being said, some preliminary research has been conducted in certain countries, For example, recent research in the US points to a potential reduction in life expectancy at birth of 1.1 years in 2020 with estimated reductions for the Black and Latino populations that are 3 to 4 times that for Whites. In the UK, recent research points to potential reductions in life expectancy at birth of 0.9 years for women and 1.2 years for men. I have not found any research in Canada thus far measuring the potential impact of the pandemic on the 2020 life expectancy. However, based on the excess deaths determined by Statistics Canada, it seems as though the impact will be lower than what is estimated in recent researches in the UK and the US.

The impact on mortality for 2021 is still highly uncertain and will depend on many factors. The rollout of Canada’s vaccination will play a major role. As you all know, Canada’s vaccination schedule has been delayed significantly so far. Although the government is confident that it will be able to make up for lost time, this remains to be seen. There are also other factors, such as the evolution of the second wave, which is still in progress, as well as potential other waves. As shown previously, the second wave in Canada has been less deadly so far, but this can change quickly, especially with the arrival of new variants of the virus. Government intervention will also play a role. Governments are trying to balance a fine line between minimizing the spread of the virus, keeping the economy going and the schools open and maintaining the mental health of the population. Their policy actions in the next months or year will have an impact on the spread of the virus. The extent to which the population will follow public health directives has also proven to be a factor in the second wave and will continue to do so as vaccinations roll out. Unlike the first wave, where the virus was mostly spreading in long-term care homes and older age groups, the virus during the second is now spreading through younger age groups at public and private gatherings.

Uncertainty in the Longer Term is even Higher (Slide 31)

Looking beyond short-term period life expectancy, the uncertainty is even higher. Considering only the direct consequences of the pandemic, and what I mean by direct is the spike in COVID-19 deaths, the impact on cohort life expectancy will depend on the underlying life expectancy of those who died from COVID-19 compared to the general population. If the individuals who died in the first wave already had low life expectancies to begin with as a result of frailties and pre-existing conditions, then the impact on cohort life expectancies will not be material. This is also referred to as the «harvesting effect » whereby a period of excess mortality is compensated by subsequent reductions in mortality. Although it is still premature to assess any potential harvesting effect, Quebec, who was hardest hit by the pandemic in the first wave did not experienced one over the summer months. If anything, there were excess deaths over the summer months in Quebec.

In addition to the direct consequences of the pandemic, we cannot underestimate the potential impact of indirect consequences. Delayed surgeries and treatments, the unknown long-term effects on the health of COVID-19 survivors, the impact of social and the isolation and job losses on mental health could all lead to increases in future mortality rates. On the other hand, behavioural changes such as social distancing, hand washing, wearing face masks, getting the flu shot can lead to decreases in future mortality rates if they are maintained in the future. As for medical research advancements, although research on certain diseases will have been delayed, other the intensive research on finding a COVID-19 vaccine may lead to breakthroughs in other diseases.

At this point, it is still too early to fully understand and estimate the impact that the pandemic will have on both short and long-term mortality rates. It will take years until the effect is known. In the meantime, actuaries and demographers will be having a lot of fun parsing out and analysing data to gain a better understanding of how this pandemic will affect future mortality patterns.

Conclusion (Slide 32)

To conclude, increasing longevity at older ages is expected to continue to put financial pressure on programs targeted to older population.

There is a lot of uncertainty related to future mortality.

In order to develop mortality assumptions, we need to understand the past, but it is even more important to look into future.

Extra care and analysis will be needed to reflect the pandemic in future mortality assumptions, especially if starting point for mortality rates is 2020 or 2021.

Thank you. I will be pleased to answer any questions you might have.