Announcement: SOA releases March 2022 Exam P passing candidate numbers.

SOA - Influenza Pandemic: The Impact on an Insured Lives Life Insurance Portfolio

Influenza Pandemic: The Impact on an Insured Lives Life Insurance Portfolio
by Dr. Andrea Stracke and Dr. Winfried Heinen

There are growing concerns that the Asian bird flu virus will evolve into a form that can be transmitted from human to human, thus unleashing a new global flu pandemic.

Flu pandemics, i.e., worldwide flu epidemics, occur at more or less regular intervals. There were three pandemics in the 20th century: the Spanish Flu pandemic in 1918/19, which caused 50 million deaths around the world (some sources even put the total up to 100 million), the Asian Flu pandemic in 1957/58, and the Hong Kong Flu pandemic in 1968/69, which resulted in around one million deaths each worldwide. The Spanish Flu alone killed approximately 4 percent of the global population at that time. It is therefore apparent that flu pandemics represent a serious threat to the population. Accordingly, the World Health Organization (WHO) called on member nations some time ago to develop plans for dealing with a pandemic.

For the insurance industry, especially life and health insurance, the effects of a pandemic would be considerable: the rise in mortality among the insured population would result in higher claims expenses for term, whole life, endowment and other life insurances paying death benefits. This would be partially offset by reduced claims expenses in annuity lines. Health insurance and disability covers would be affected as well. In addition, a pandemic is likely to have a major impact on capital markets, which would have implications for the assets side of insurance companies' balance sheets.

We will limit our discussion here to an analysis of the additional claims expenditure in a (German) life insurance portfolio due to a pandemic.

It is impossible to reliably predict the outbreak or effects of a new pandemic: On the one hand, medical care has improved since 1918/19; on the other, the increased mobility of today's population would cause the virus to spread much faster. We therefore decided to re–model the historic Spanish Flu event based on today's population structure and in a manner that would enable the analysis of a wide range of scenarios. We leave it to the reader to assess which of the presented scenarios carries the highest probability.

Spanish Flu
The Spanish Flu spread in three waves (in the spring and autumn of 1918 and in early 1919), with the second wave being the most aggressive. Compared to other pandemics, the Spanish Flu was not exceptionally contagious: infection rates of around 30 percent per wave are common for a pandemic. There were other drivers for the deadliness of the Spanish Flu:

  • A critical factor was the high case fatality rate in comparison with normal waves of flu.
  • The mortality pattern by age group during that pandemic was unusual. Flu mortality by age usually has a "U" shape–the rate is highest among children and the elderly. The curve formed by the Spanish Flu, however, resembled a "W" with a peak in the middle representing young adults.

The brief interval between the three waves of the pandemic was atypical as well.

The reasons for the unique features of the Spanish Flu remain unclear and are a matter of dispute in the pertinent literature.

The Spanish Flu is generally portrayed as a worst–case scenario. According to some theories, however, the Spanish Flu virus was related to a strain that had arisen before 1889 (cf. Taubenberger, Morens [7]). In that case, a portion of the older population in 1918 would have been immune to the Spanish Flu virus. Without that degree of immunity among the older population, the effects of the Spanish Flu could have been worse (cf. Willets [9]).

Regarding the current bird flu, it is claimed that around 50 percent of all cases reported to date have been fatal. This mortality rate is several times higher than the rate for the Spanish Flu. However, to trigger a pandemic the bird flu virus would have to mutate, and it is assumed that this would change its degree of aggressiveness. Moreover, it can be assumed that only the most serious cases are being reported at present and that minor cases of bird flu among humans may be going undetected. If so, the 50 percent case fatality rate would be too high. It should be noted that the SARS virus, which is transmittable between humans, had a case fatality rate of 9.6 percent (cf. WHO [10]).

A simple estimate for the claims costs incurred by an insurance portfolio due to a pandemic can be made by multiplying the portfolio's total sum at risk by the pandemic's average (and thus not age–specific) mortality rate for that portfolio.

Most of the pertinent literature deals with the increased mortality rate among the general population rather than among insured lives. To derive a figure for the latter group, we divided in our model the overall mortality rate ascribed to the pandemic into the components "infection rate" and "case fatality rate," for which we derived age band– and risk group–specific probabilities. The higher level of detail then enabled us to examine scenarios of varied degrees of severity with regard to infection and/or case fatality rates.

Age Groups
Both infection and case fatality rates vary strongly by age: the infection rate tends to be highest among schoolchildren and lowest among the elderly population:

  • First of all, older citizens generally have less social contact.
  • Moreover, they have been exposed to many related flu viruses in the past and are therefore partially immune to the latest strain.

As explained above, flu mortality rates usually have a "U" or "W" shape. Consequently, we modeled the effects of a pandemic by age bands.

The most comprehensive data for modeling the infection rates and pandemic mortality were compiled in the United States, which was apparently hit harder by the Spanish Flu than Germany: Niall et al. [4] reported a rise in the general mortality rate of 0.65 percent in the United States compared to only 0.38 percent in Germany. The findings of the various American studies are inconsistent, however. We were, therefore, forced to gather data from different studies and combine it for our model.

Deriving case fatality rates was particularly problematic. In general, only the additional mortality rate among the general population is reported and analyzed. To derive case fatality rates, this figure must then be divided by the infection rates.

The mortality rate among infected with pre–existing conditions such as respiratory diseases is considerably higher than among those who were healthy at the time of infection. The relevant risk profiles for Germany and the United States differ, which means that age–specific mortality rates in the United States cannot be directly applied to Germany without modification. Like Meltzer et al. [2] and in the Swiss study [5], we therefore included the state of health prior to infection ("healthy" or "pre–existing condition") in the calculation of age–specific case fatality rates. We then applied the German risk profile to convert the age– and risk–group specific (American) mortality rates into age–band (but not risk–group) specific German case fatality rates.

In contrast to the case fatality rates, we assumed that infection rates were not linked to the victims' state of health, for which reason the modification of the data was not necessary in this case.

Using the American data, we modeled the effects of the Spanish Flu outbreak in the United States in 1918/19 and applied the results to the age distribution of today's German population. We used the 2002 data contained in the 2004 Statistical Yearbook to determine the German population structure. For the breakdown of the data per age–band into the categories "healthy" and "pre–existing condition," we applied to the German population the relevant percentages of the Swiss population used in the Swiss study [5].

We modeled the pandemic in three waves. As stated previously, the second wave was the most aggressive and thus caused higher rates of infection and case fatality than the other two. We had no data on the relation between the first and third waves and therefore assumed identical rates of infection and case fatality. In addition, we assumed that the members of the infected population who survived one wave were immune to the next. This corresponds to the methods used in the Swiss study [5]. We applied the age–specific data from that study and the study by Glezen [1] to determine the infection rates during the first and third waves (see Table 1).

For today's German population, the figures result in an overall infection rate of 25 percent for the first wave. For the second wave, in line with the Swiss study, we estimated an infection rate of 40 percent of the population not infected during the first wave, and assumed an age distribution in accordance with the distribution of the infection rates during the first wave. For the third wave, the infection rates in Table 1 were used and applied to the non–immunized population. This results in a total infection rate of around two–thirds of the population for all three waves.

Table 2 shows the overall age–specific flu–related population mortality rates for the years 1918 and 1919; cf. [8]. We assumed that the deaths in 1919 were related to the third wave and the deaths in 1918 to waves one and two. Proceeding from our previous assumption that the mortality rate was the same for the first and the third wave, we calculated that 38 percent (= 223 / 588.5; cf. Table 2) of the increased mortality rate in 1918 was attributable to the first wave and 62 percent to the second. However, not the entire population was at risk during the second wave due to the fact that the infected survivors of the first wave were immune. To compute the mortality rate of the exposed population for the second wave, we had to divide the 62 percent of the total additional mortality in 1918 by the proportion of the population not infected during the first wave, which Glezen [1] put at 70.6 percent. Finally, all three mortality rates thus derived were divided by the corresponding infection rates, in which regard we used for the second wave the 40 percent average infection rate estimated above. See Table 3.

Flu Medicine, Vaccinations and Other Measures
If a pandemic hits in several waves, it can be assumed that a vaccination will have been developed by the second or third wave at the latest. This would be of little benefit, however, to the majority of insureds in a life insurance portfolio as would be flu medicines such as the much–discussed Tamiflu, because enough vaccine or flu medicine would not be available for the entire population. We therefore decided not to take the effects of flu vaccinations and treatments into account for this article, although they could have been implemented in our model. For this purpose we also refrained from including the possible effects of other epidemic–prevention measures such as school closings and quarantines.

Based on the information presented, we can now venture to predict the effects of various pandemic scenarios on the mortality rate of the overall population or of a life insurance portfolio. In the following, differences between the two rates are only attributable to different age structures. The fact that the distribution of risk groups in an insured lives portfolio is more favorable than in the overall population has been disregarded here.

We are looking at a model insured lives portfolio in which insureds are aged 15 to 65 and the distribution of age groups corresponds to that of the German population. Given that insureds are presumably older on average than the German population in the age group 15 to 65, the risk of infection declines with age and the case fatality rates only begin to rise significantly after age 66, our assumption is conservative for our purposes.

The additional claims frequency triggered by a pandemic in various scenarios is illustrated in Table 4.

As the figures show, the additional mortality both in the general population and among insured lives varies proportionally to the case fatality. Contrary to what may be initially assumed, this does not hold true for the infection rate. This is attributable to the modeling of the pandemic in three waves and the resulting immunization effects.

How would the increase in insured mortality impact claims expenses in Germany? According to the latest statistics compiled by the German Insurance Association (GDV), the total sum insured under life insurance policies in the German market amounted to €1,626 billion at the end of 2004. We assume that the associated sum at risk totals around €1,000 billion. Multiplied by 0.46 percent, the pandemic mortality of an insurance portfolio, this equates to net claims expenses of €4.6 billion for the market as a whole. Depending on the scenario, this figure fluctuates between €2.3 billion and €9.1 billion, even soaring to €43.5 billion for the SARS scenario.

Additional claims costs of nearly €5 billion–around 50 percent of the market's total annual gross profit (before policyholder bonuses)–would strain, but surely not break, the German insurance market. The same would hold true for a more pessimistic scenario with doubled loss costs. In the worst–case scenario of a flu wave with SARS–like case fatality rates, however, the additional claims expenses would reach around €45 billion; this would equal five times the market's total annual gross profit or 100 percent of the market's policyholder bonus reserves. In principal, even this added burden could be absorbed by the insurance industry without jeopardizing the companies' ability to honor long–term commitments. Nevertheless, sustaining a hit of this magnitude to the balance sheet in just one year would be a difficult task. In that type of situation–the death of 6 percent of the population within a short period of time with the resulting impact on the economy as a whole–it would be legitimate, at least, to consider regulatory measures.

As stated several times in this article, our model is capable of generating other, more detailed scenarios that cannot be discussed here. Instead, a concluding word on our own behalf: in contrast to other extreme scenarios, such as regional natural disasters, the advantages that reinsurers have over primary insurers due to geographical diversification cease to apply in the event of a pandemic. Consequently, the net risk–bearing capacity of a reinsurer per region is lower in the latter case than for other catastrophic events. A primary insurer wishing to buy reinsurance protection against the risk of a pandemic should therefore take care to ensure–even more so than when covering other catastrophe scenarios–that its reinsurer either has its pandemic exposure well under control or has a very strong capital base.

Dr. Andrea Stracke is a reinsurance actuary for Gen Re LifeHealth. She can be reached at

Dr. Winfried Heinen is chief actuary for the international business of Gen Re LifeHealth. He can be reached at

Reprinted with permission from Assets & Liabilities, A Gen Re Publication.


  1. Glezen, W. Paul: Emerging Infections: Pandemic Influenza. Epidemioplogic Reviews. Vol. 18, No. 1, 1996.
  2. Meltzer, Martin I., Cox, Nancy J. and Fukuda, Keiji: The economic impact of pandemic influenza in the United States: priorities for intervention. Emerging Infectious Diseases, 1999; 5(5): 659-671.
  3. Nationaler Influenzapandemieplan Teil II–Analysen und Konzepte für Deutschland. Juli 2005, Bericht der Expertengruppe "Influenza–Pandemieplanung" am Robert Koch–Institut.
  4. Niall, P., Johnson, S., and Mueller, Juergen: Updating the Accounts: Global Mortality of the 1918-1920 "Spanish" Influenza Pandemic. Bull. Hist. Med., 2002, 76:105-115.
  5. Piercy, James and Miles, Adrian: The Economics of Pandemic Influenza in Switzerland. March 2003.
  6. Statistisches Jahrbuch 2004, Statistisches Bundesamt.
  7. Taubenberger, Jeffrey K., Morens, David M: 1918 Influenza: The Mother of All Pandemics. Emerging Infectious Diseases, Vol. 12, No. 1, January 2006.
  8. The Pandemic of Influenza in 1918-1919. U.S. Department of Health, Education, and Welfare. Public Health Service. National Office of Vital Statistics. (
  9. Willets, Richard: The facts behind the hype. The Actuary, December 2005.
  10. World Health Organization (WHO): Summary of probable SARS cases with onset of illness from 1 November 2002 to 31 July 2003. (