Subject CS2 Revision Notes For the 2021 exams Exposed to risk Booklet 5 covering Chapter 9 Exposed to risk The Actuarial Education CompanyPAST EXAM QUESTIONS This section contains all of the past exam questions from 2010 to 2018, plus those from the 2019 Paper A exams, relating to the topics covered in this booklet. Solutions are given after the questions. These give enough information for you to check your answer, including working, and also show you what an outline examination answer should look like. Further information may be available in the Examiners’ Report, ASET or Course Notes. (ASET can be ordered from ActEd.) We first provide you with a cross-reference grid that indicates the main subject areas of each exam question. You can use this, if you wish, to select the questions that relate just to those aspects of the topic that you may be particularly interested in reviewing. Alternatively, you can choose to ignore the grid, and attempt each question without having any clues as to its content. Page 12 © IFE: 2021 ExaminationsCross-reference grid paydwaye uoysenp ‘Saydde ayer uy YoIUM 0} aby ‘soyes Buyewnsy pouyeur snsus0— Y19 woRenaeo pexe- 413 ‘SOUapPUddSOLIOD 30 @dioung eyep Bulpiipqns, AyouSHOIS OY jo walqold uonsend 10 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Page 13 © IFE: 2021 ExaminationsSubject CT4 April 2010 Question 1 List four factors often used to subdivide life insurance mortality statistics. [2] Subject CT4 April 2010 Question 6 An oil company has discovered a vast deposit of oil in an equatorial swamp. The area is extremely unhealthy and inhabited by venomous spiders. There is an antidote to bites from these spiders but it is expensive. The antidote acts instantly but does not provide future immunity. The company commissions a study to estimate the rate of being bitten by the spiders among its employees, in order to determine the amount of antidote to provide Employees of the company are posted to the swamp for six month tours of duty starting on 1 January, 1 April, 1 July or 1 October. The first employees to be posted arrived on 1 January 2008. The swamp is so inaccessible that no employees are allowed to leave before their six month tours of duty are completed. Accidental deaths are common in this dangerous location. The table below gives some data from the study. Quarter Number of new Number of Number of beginning arrivals at start accidental spider bites of quarter deaths during during quarter quarter 1 January 2008 90 10 15 1 April 2008 80 8 25 4 July 2008 114 10 30 1 October 2008 126 13 40 () Estimate the quarterly rate of being bitten by a spider for each quarter of 2008, stating any assumptions you make. 7 (i) Suggest reasons why the assumptions you made in (i) might not be valid (1) [Total 8] Page 14 © IFE: 2021 ExaminationsSubject CT4 September 2010 Question 7 Two neighbouring small countries have for many years taken annual censuses of their populations on 1 January in which each inhabitant must give his or her age. Country A uses an ‘age last birthday’ definition of age, whereas Country B uses an ‘age nearest birthday’ definition. Each country has also operated a system in which deaths are recorded on an ‘age nearest birthday at date of death’ basis. On 30 June 2009 Country A invaded Country B and the two countries became one state. The new government wishes to estimate a single set of age-specific death rates, 1, , for the new unified state using the census data taken in the years before the invasion. Derive a formula which the new government may use to estimate ju, in terms of the recorded number of deaths in each country, and the population of each country recorded as being aged x in the censuses. State any assumptions you make. [8] Subject CT4 September 2011 Question 9 (i) State the principle of correspondence as it applies to the estimation of mortality rates. 11] (ii) Explain why it might be difficult to ensure the principle of correspondence is adhered to, and give a specific example of an investigation where this may be the case. 2) An actuary was asked to investigate the mortality of lives in a particular geographical area. Data are available of the population of this area, Classified by age last birthday, on 1 January in each year. Data on the number of deaths in this area in each calendar year, classified by age nearest birthday at death, are also available. (ii) Derive a formula which would allow the actuary to estimate the force of mortality at age x +f, ,,/, ina particular calendar year, in terms of the available data, and derive a value for f [6] (iv) List four factors other than geographical location which a government statistical office might use to subdivide data for national mortality analysis. [2] [Total 11] © IFE: 2021 Examinations Page 15Subject CT4 April 2012 Question 2 () Explain the reasons why data are subdivided when conducting mortality investigations. (ji) Describe the problems which can arise with subdividing data [2] [Total 4] Subject CT4 April 2012 Question 9 () Uist four factors other than age and smoker status by which life insurance mortality statistics are often subdivided. 2] ‘Two offices in different towns of the same life insurance company write 25-year term assurance policies. Below are data from these two offices relating to policyholders of the same age. Both deaths and policies in force are on an age last birthday basis. Gasperton | Great Hawking Policies in force on 1 January 2009 2,000 1,770 Policies in force on 1 January 2010 2,100 1,674 Deaths in calendar year 2009 25 21 (i) Calculate the central death rate for the calendar year 2009 at this age for the offices in Gasperton and Great Hawking. 2] A detailed examination of the records shows that 50% of the policyholders in Gasperton at both censuses were smokers, and 20% of policyholders in Great Hawking at both censuses were smokers. National death rates at this age for smokers in 2009 were 40% higher than those for non-smokers. (ii) Estimate the central death rates for smokers and non-smokers in Gasperton and Great Hawking [4] The life insurance company charges policyholders in Gasperton and Great Hawking the same premiums for the 25-year term assurance policies. It charges smokers in both towns 40% more than non-smokers. (iv) Comment on the company’s pricing structure in the light of your results from parts (ii) and (iii) above. [3] [Total 11] Page 16 © IFE: 2021 Examinations7 Subject CT4 September 2012 Question 3 (i) State the principle of correspondence as it applies to mortality rates. [1] Alife insurance company has the following data: ‘Number of policies in force on Age last January ‘January 1July 1 January birthday 2009 2010 2010 2011 49 2,000 2,100 2,300 2,500 50 2,100 2,200 2,300 2,400 51 2,300 2,400 2,500 2,600 Number of deaths classified by age next birthday and calendar year Age next birthday 2009 2010 49 175 200 50 200 225 51 225 235 (ii) Estimate, using these data, the force of mortality at age 50 next birthday for the period 1 January 2009 to 1 January 2011. (5) (iii) State the exact age to which your answer to part (ii) relates. tu) [Total 7] © IFE: 2021 Examinations Page 17Subject CT4 April 2013 Question 5 Population censuses in a certain country are taken each year on the President's birthday, provided that the President's astrological advisor deems the taking of a census favourable. Censuses record the age of every inhabitant in completed years (that is, curtate age). Deaths in this country are registered as they happen, and classified according to age nearest birthday at the time of death Below are some data from the three most recent censuses. completed Population 2006 | Population 2009 | Population 2010 ions (thousands) (thousands) (thousands) 64 300 320 350 65 290 310 330 66 280 300 320 Between the censuses of 2006 and 2009 there were a total of 3,000 deaths to inhabitants aged 65 nearest birthday, and between the censuses of 2009 and 2010 there were a total of 1,000 deaths to inhabitants aged 65 nearest birthday. () Estimate, stating any assumptions you make, the death rate at age 65 years for each of the following periods: * the period between the 2006 and 2009 censuses + the period between the 2009 and 2010 censuses. 16] (i) Explain the exact age to which your estimates apply. (1) [Total 7] Subject CT4 September 2013 Question 1 Data are often subdivided when investigating mortality statistics. () Explain why this is done. [2] (ii) Discuss one potential problem with subdividing mortality data. (2) (iii) List four factors which are commonly used to subdivide mortality data. [2] [Total 6] Page 18 © IFE: 2021 Examinations10 1 Subject CT4 April 2014 Question 2 () Explain why data are subdivided into homogeneous groups when mortality investigations are conducted. 2 (ii) List four factors, other than age and sex, by which mortality statistics are often subdivided 2) [Total 4] Subject CT4 April 2014 Question 4 () State the principle of correspondence as it relates to mortality investigations. (1) Two small countries conduct population censuses on an annual basis. Country A records its population on 1 February every year based on an age definition of age last birthday. Country B records its population on every 1 August using a definition of age nearest birthday. Each country records deaths as they happen based on age next birthday. Below are some data from the last few years. Country A Age last Population Population Population birthday 1 February 2011 1 February 20121 February 2013 44 382,000 394,000 401,000 45 374,000 381,000 385,000 46 354,000 372,000 375,000 Country B Age nearest Population Population Population birthday 1 August 2011 1 August 2012 1 August 2013 44 382,000 394,000 401,000 45 374,000 381,000 385,000 46 354,000 372,000 375,000 In the combined lands of Countries A and B in the calendar year 2012 there were 4,800 deaths of those aged 46 next birthday and 4,500 deaths of those aged 45 next birthday. The two countries decide to form an economic union, after which it will be mandatory to offer the same rates for life insurance to residents of each country, © IFE: 2021 Examinations Page 1912 (ii) Estimate the death rate at age 45 years last birthday for the two countries combined. [6] (iii). Explain the exact age to which your estimate relates. tu [Total 8] Subject CT4 September 2014 Question 3 @ Explain the census approximation for calculating the exposed to risk between any two census dates. [2] A mortality investigation bureau has collected the following information on number of policies in-force each year from different companies. Wee Year | CompanyA | CompanyB | Company C ‘Age 54 | 2011 3.400 1,250 5,780 2012 3,350 1,450 5,500 2013 3,000 1.500 6,010 [Age 55 | 2011 3.250 1,190 6,000, 2012 3,390 1.300 5,960 2013 3,100 1,440 6,030 Age 56 | 2011 3.270 1.150 5.950 2012 3,020 1,300 5,980 2013 2,950 1,500 5,990 * Company A has provided in-force policy data as at the beginning of each calendar year using age nearest birthday. Company B has provided in-force policy data as at the financial year closing date (which was 31 March in each year) using age last birthday. Company C has provided in-force policy data as at the end of each calendar year using age next birthday. (i) Calculate the contribution to central exposed to risk for lives aged 55 last birthday for the calendar year 2012 for each of the companies. [6] [Total 8] Page 20 © IFE: 2021 Examinations13 14 Subject CT4 April 2015 Question 5 (State the principle of correspondence as it applies to death rates. [1] Anightclub opens at 10:00pm and closes at 2:00am. It admits only people aged over 21 years on the production of an identity card giving date of birth. The table below shows the number of people entering in various intervals between 10:00pm and 2:00am on 30 June 2013. No one was admitted after 1:00am, and you may assume that all those who enter the premises stay until 2:00am Year of birth 10:00— 11:30pm = 11:30 — 12:00pm 12:00pm — 1:00am 1989 100 300 200 1990 200 400 350 1991 150 400 300 1992 100 250 200 During the period of opening, 40 people aged 22 last birthday required medical attention for heat exhaustion. (ii) Calculate the rate per person-hour at which those attending the nightclub aged 22 last birthday required medical attention for heat exhaustion, stating any assumptions you make. (6) [Total 7} Subject CT4 April 2015 Question 8 (i) State why it is important to divide data into homogeneous classes when undertaking mortality investigations. 2] (ii) List four factors, apart from smoking behaviour, by which mortality data are often classified by life insurance companies. [2] In a particular life insurance market, it has for many years been the practice for all companies to charge smokers higher premiums than non-smokers for the same term assurance policy. Suppose one company decides to switch to charging smokers and non-smokers the same premiums for term assurance policies. The other companies retain differential pricing for smokers and non-smokers. (iii). Discuss the likely implications for the company making the switch. _[4] [Total 8] © IFE: 2021 Examinations Page 2115 16 Subject CT4 September 2015 Question 1 List four factors, other than age and sex, by which mortality statistics are often subdivided. Subject CT4 September 2015 Question 4 Company A and Company B are two small insurance companies which have recently merged to form Company C. Company C is reviewing its premium rates for a whole of life product and so is conducting an analysis of mortality rates experienced. ‘Company A recorded the number of policies in force every 1 January using a definition of age next birthday whereas Company B recorded the number of policies in force every 1 April using an age definition of age last birthday. Both companies recorded deaths as they happened using an age definition of age last birthday. These are the data for the most recent years. Company A Age next ‘Number of Number of ‘Number of birthday —_ policies 1 Jan policies 1 Jan policies 1 Jan 2012 2013 2014 51 8,192 6,421 8,118 52 7,684 8,298 7,187 53 9,421 8,016 9,026 Company B Age last Number of Number of Number of birthday policies 1 April _— policies 1 April _ policies 1 April 2012 2013 2014 51 4,496 3,817 4,872 52 5,281 5,218 3,812 53 4,992 5,076 5,076 In the calendar year 2013 Company A recorded 28 deaths of those aged 52 last birthday and Company B recorded 17 deaths of those aged 52 last birthday. Page 22 © IFE: 2021 Examinations17 18 (i) Estimate the force of mortality for the combined company for age 52 last birthday, stating all assumptions that you make. (6) (ii) Explain the exact age to which your estimate applies. t) [Total 7} Subject CT4 April 2016 Question 1 Write down the information required to compute the exact exposed to risk in an investigation of mortality. 3] Subject CT4 April 2016 Question 3 (i) State the principle of correspondence in the context of a mortality investigation. (1) A mortality investigation collects the following data: 1n,(t) = total number of policies under which death claims are made when the policyholder is aged x last birthday for each calendar year t . P, (t) = number of in-force policies where the policyholder was aged x nearest birthday on 1 January in year ¢. (i) (@) Derive an expression, in terms of P,(t) , for the central exposed to risk, E¢, corresponding to the claims data which may be used to estimate the force of mortality in year t ateach age x, sy (b) State any assumptions that you make, indicating at which point in your derivation each assumption is relevant (5) [Total 6] © IFE: 2021 Examinations Page 2319 Subject CT4 September 2016 Question 8 An analysis of the number of term assurance policies in force for three ‘companies has revealed the following information: Year Company A Company B Company C 2013 6,728 2,643 4,132 Age 50 2014 6,189 2,548 2015 5,962 2,496 4,630 2013 5,987 2,333 4,012 Age 51 2014 6,002 2,417 2015 5,056 2,213 4,500 2013 5,359 2,155 3,895 Age 52 2014 5,600 1,992 2015 4,906 2,006 4,367 * Company A has reported the number of policies in force on 1 January each year using age nearest birthday, * Company B has reported the number of policies in force on 1 November each year using age last birthday. * Company C has reported the number of policies in force on 31 December each year using age next birthday, but failed to provide data for 2014. () Calculate the contribution to the central exposed to risk for lives age 51 last birthday for the calendar year 2014 for each company individually. [5] (i) (a) State the assumptions you have made in order to perform your calculations. (b) Explain why these assumptions were required. [6] [Total 11] Page 24 © IFE: 2021 Examinations20 Subject CT4 April 2017 Question 11 A large company operates a health benefits scheme which pays a sickness benefit to any employee who is unable to work through ill-health and a death benefit to any employee who dies. (i) Drawa transition diagram with three states which could be used to analyse data from this scheme. (2) (i) Give the likelihood of the data, defining all the terms you use. [4] (iii) Derive the maximum likelihood estimator of the rate of falling sick. [3] The company records data on 1 January each year, classified by age last birthday on: * the total number of employees (including those in receipt of sickness benefit). * the number of employees in receipt of a sickness benefit. ‘Some recent data are given in the table below: Age last Total number of employees on Number of employees in birthday 1 January in year receipt of sickness benefit on 1 January in year 2014 2015 2016 2014 2015 2016 51 148 162 180 12 20 8 52 146 148 160 10 18 7 53 140 144 146 8 20 6 The company wishes to estimate the rates of falling sick and recovery at age 52 years nearest birthday over the two-year period consisting of the calendar years 2014 and 2015. (iv) Determine suitable exposed-to-risks for calculating the rates of falling sick and recovery. [Total 15] © IFE: 2021 Examinations Page 252 Subject CT4 September 2017 Question 4 Astudy was conducted into the mortality of persons aged between exact ages 85 and 86 years. The study took place from 1 April 2015 to 31 March 2016. The following table shows information on 10 lives observed in the study. Life number Date of 85th birthday Date of death 1 1 August 2014 - 2 1 November 2014 - 3 1 January 2015 1 February 2016 4 1 February 2015 - 5 1 March 2015 - 6 1 April 2015 1 January 2016 7 1 June 2015 1 November 2015 8 1 July 2015 - 9 1 September 2015 1 March 2016 10 1 January 2016 7 () Calculate a central exposed to risk for the 10 lives in the sample, working in months. (3) (i) Give the maximum likelihood estimate of the mortality hazard at age 85 last birthday. (1) (ii) Estimate qs 1] [Total 5] Page 26 © IFE: 2021 Examinations22 Subject CT4 April 2018 Question 6 The National Statistics Office of a smalll, low income country wants to estimate recent death rates. A death registration system has allowed the National Statistics Office to estimate deaths by age nearest birthday for the ten-year period 1 January 2005 - 31 December 2014. Censuses of this country are infrequent. A successful census was completed on 1 January 2015, but the previous reliable census was on 1 January 2002. Both censuses collected data on the population aged x last birthday by single years of age. () Explain why the National Statistics Office should adjust the age definition in the census data to correspond with that of the deaths data. (2) Let the census population at age x last birthday on 1 January in year t be Pyy. (ii) Derive an expression, in terms of the P,; , for the exposed to risk for the period covering the years 2005 to 2014 inclusive which the National Statistics Office could use to estimate the overall death rate at age x nearest birthday. [5] (iii) Set out any assumptions you make in your derivation in part (i), indicating where in the derivation they are needed 2) The death registration system in this country is being maintained, but the next census is not planned until 2025. (iv) Discuss how you might estimate death rates at age x nearest birthday for the calendar years 2015 and 2016. 13] [Total 12] © IFE: 2021 Examinations Page 2723 Subject CT4 September 2018 Question 10 A religious organisation maintains two lists of members: * alist of sick members, so that members may pray for them (the Sick List); * alist of recently deceased members (the Dead List). Each list is published in a bulletin given to those attending the regular weekly meetings of religious worship. The lists are updated each week half way between the religious worship meetings. A study was made of the mortality of sick members. A sample of members joining the Sick List in the first quarter of 2016 was followed until they left the list. Those who left the list but who did not move to the Dead List were assumed to have recovered. The study terminated on 31 March 2017. Below are given some data from the study. ‘Week first appeared on Sick List’ and ‘Week last appeared on Sick List’ are measured in weeks from the first week of 2016. Week first Week last —— appeared on appeared on Outcome Sick List Sick List i i i ‘Assumed recovered 2 1 3 Moved to Dead List 3 3 4 Moved to Dead List Still on Sick List : 2 S 31 March 2017 5 6 17 Moved to Dead List 6 7 14 ‘Assumed recovered 7 9 1 ‘Assumed recovered 8 10 60. Moved to Dead List 9 11 11 Moved to Dead List Still on Sick List 7 12 65 31 March 2017 () Calculate the central death rate per week of these members using the exact exposed to risk. [4] (i) Determine the probability of survival for 52 weeks using your result from part (i). (1) Page 28 © IFE: 2021 ExaminationsOne member of the organisation has been studying statistics and recommends using the Nelson-Aalen estimator to calculate the probability of survival for 52 weeks. (iil) Calculate the Nelson-Aalen estimate of S(52) . 5) (iv) Comment on your results in parts (ii) and (ii). 2] [Total 12] Subject CT4 September 2018 Question 11 Man Life is an insurance company which only sells life insurance to males. It has recently bought another smaller company called Mixed Life which sells business to both males and females. The company is reviewing the premium rates it charges for life insurance. Man Life has records of the number of policies in force at their year end, which is 30 September, recorded by age last birthday. Mixed Life has records of the number of policies in force on 31 December each year recorded by age last birthday for males and age nearest birthday for females. These are the data for the most recent years. Man Life Age last | Number of policies | Number of policies | Number of policies birthday 30 Sept 2015 30 Sept 2016 30 Sept 2017 49 4,789 4,296 4,367 50 4,953 5,009 4,809 51 5,300 5,186 5,902 Mixed Life Males Age last | Number of policies | Number of policies | Number of policies birthday 31 Dec 2015 31 Dec 2016 31 Dec 2017 49 1,832 1,650 1,698 50 1,800 1,750 1,550 51 1,966 1,756 1,569 © IFE: 2021 Examinations Page 29Mixed Life Females Age last | Number of policies | Number of policies | Number of policies birthday 31 Dec 2015 31 Dec 2016 31 Dec 2017 49 1,602 1,568 1,639 50 1,506 1,497 1,508 51 1,610 1,587 1,411 (@ Calculate the central exposed to risk of the combined portfolio for males aged 50 last birthday for the calendar year 2016, stating each assumption you make at the point where you make it. [5] (i) Calculate the central exposed to risk of the combined male and female Portfolio for persons aged 50 last birthday for the calendar year 2016, stating each assumption you make at the point where you make it. [3] Legislation has been brought in which means that males and females must be charged the same premium rates for life insurance. The company is considering basing its future premium rates on the number of deaths across the whole male and female portfolio of the two companies at each age divided by the exposed to risk across the combined male and female portfolio at each age. (ii) Discuss the appropriateness of the company's approach to determining its future premium rates. (3) [Total 11] Subject CS2 April 2019 Question 1 State why data are divided into homogeneous groups when conducting mortality investigations. (2) Page 30 © IFE: 2021 Examinations26 Subject CS2 April 2019 Question 5 () State the principle of correspondence as it applies to death rates. [1] A country uses an administrative data system which shows the estimated population aged x last birthday on 1 January each year. Deaths are registered on an age nearest birthday basis. (ii) Derive a formula which may be used to estimate an appropriate exposed to risk for calculating the average death rate at age x exact in the two-year period 1 January 2016 - 1 January 2018, defining all the terms you use. Assume that birthdays are evenly distributed across the calendar year. 4] Someone remarks that, in this country, only one third of births take place in the first half of the calendar year. (iii) Discuss the implications of this for the formula you have derived in part (ii). [3] [Total 8] © IFE: 2021 Examinations Page 31SOLUTIONS TO PAST EXAM QUESTIONS The solutions presented here are just outline solutions for you to use to check your answers. See ASET for full solutions. Subject CT4 April 2010 Question 1 Mortality statistics are often subdivided by: * age . sex © postcode * policy type. Also see Core Reading Paragraphs 3 and 4. Subject CT4 April 2010 Question 6 () Estimate the quarterly rate We will assume that all employees who are bitten by spiders are given the antidote and make an immediate recovery and that the only deaths that occur are accidental deaths. First quarter The first quarter started on 1 January with 90 employees in the first cohort and ended with 80 (= 90-10) on 31 March. If we assume that the deaths ‘occurred uniformly over the quarter, this gives a total employee exposed-to- risk for this period of (90 +80) = 85 quarters. Since there were 15 spider bites during this period, the rate of spider bites is 15 —=171 fe ae = 17.8% per quarter Page 32 © IFE: 2021 Examinations‘Second quarter The second quarter started on 1 April with the remaining 80 employees from the first cohort plus the 80 new arrivals in the second cohort, making a total of 160, and ended with 152 (= 160-8 ) on 30 June. If we assume that the deaths occurred uniformly over the quarter, this gives a total employee exposed-to-risk for the second quarter of 4 (160 +152) = 156 quarters. At the start of the second quarter there were equal numbers (80 of each) of employees from the first and second cohorts. If we assume that the proportion of deaths during each quarter was the same for each cohort, we would expect that the 8 deaths during the second quarter were equally split between the two groups. So 4 of the employees from the first cohort would go home on 30 June and 4 employees from the second cohort would have been killed. Third quarter Based on the assumption of equal proportions of deaths for each cohort, the third quarter started on 1 July with the remaining 76 (= 80-4 ) employees from the second cohort plus the 114 new arrivals from the third cohort, making a total of 190, and ended with 180 (= 190-10) on 30 September. This gives a total employee exposed-to-risk for the third quarter of 4(190 +180) = 185 quarters. Again, if we assume that the deaths during the third quarter were in proportion to the numbers at the start of the quarter, then the 10 deaths during the third quarter would consist of g *10=4 from the second cohort and 6 from the third cohort. Fourth quarter The fourth quarter started on 1 October with the remaining 108 (=114-6) employees from the third cohort plus the 126 new arrivals from the fourth cohort, making a total of 234, and ended with 221 (= 234-13) on 31 December. This gives a total employee exposed-to-risk for the fourth quarter of 3 (234 +221) = 227.5 quarters. © IFE: 2021 Examinations Page 33Again, if we assume that the deaths during the fourth quarter were in Proportion to the numbers at the start of the quarter, then the 13 deaths during the fourth quarter would consist of on *13=6 from the second cohort and 7 from the third cohort. The results of these calculations are summarised in the following table: Quarter Number Number Central Spider Estimated at start atend ETR bites rate 1st 90 80 85 15 17.6% 2nd 160 152 156 25 16.0% 3rd 190 180 185 30 16.2% 4th 234 221 227.5 40 17.6% (ji) Why the assumption might not be valid We have assumed that the probability of dying does not vary by duration, the length of time the employees have been in the region. However, we might expect that each cohort of employees will become more experienced by the second half of their tour and will be less likely to be killed. For the same reason, the assumption of uniform deaths over each quarter may not be valid. ‘The question only considers accidental deaths. However, there may also be deaths from other causes, eg tropical diseases. Page 34 © IFE: 2021 ExaminationsSubject CT4 September 2010 Question 7 Period of investigation Suppose that: + the period of investigation covers an n-year period and starts on a 1st January, * time is measured in years since the start of the investigation. Death data Death data are classified according to age nearest birthday for both countries. Let: + df =the number of deaths aged x nearest birthday in Country A + d® =the number of deaths aged x nearest birthday in Country B ¢ d,=d4+d° = total number of deaths aged x nearest birthday. Census data The census data are classified according to age last birthday for Country A and age nearest birthday for Country B. So the census data and death data match for Country B but not for Country A. Exposed to risk for Country A Let PA(t) denote the number of lives in Country A at time t aged x last birthday. The values of P;* (0), P(1),....A(n) are known for all x . However, since the census data don't match the death data, we define another function PZ" (t) to be the number of lives in Country A at time t aged x nearest birthday. (This function does match the death data.) Then the central exposed to risk for Country A is: Ef = (SPX (that © IFE: 2021 Examinations Page 35Assuming that P;“"(t) varies linearly between the census dates: EGA = ST PR (0) + Pe (tee 3] PA (1) +26" (n)] = EE [Ae + PA (k+1)] Assuming that birthdays are uniformly distributed over the calendar year: ° 1 PE (t) = [PAH +PA (0) for all times tf. So: es Ela A, (k) + PA (k) +PAs(k +1) +PA(k+4)] Exposed to risk for Country B Let P2(t) denote the number of lives in Country B at time t aged x last birthday. The values of P(0),P2(1),....P8(n) are known for all x and the central exposed to risk for Country B is given by: Ep? = [ge (that Assuming that P2(t) varies linearly between the census dates, we have: ech. 15 [Pa (&) +P2(k+1)] P= DFE " Total exposed to risk The combined central exposed to risk for the unified state is then: = EGA ES Page 36 © IFE: 2021 ExaminationsEstimate of the force of mortality ‘Assuming that the force of mortality is constant over the rate interval “x nearest birthday’, it can be estimated by: The rate interval starts at exact age x-% and ends at exact age x +%, so the average age of the rate interval is x. Hence ji, is an estimate of 41, Subject CT4 September 2011 Question 9 (i) Principle of correspondence The principle of correspondence states that a particular day should be included in the exposed to risk for an individual life if and only if, were the person to die on that day, they would be included in the corresponding death count. (ii) Difficulties of ensuring the principle of correspondence The main reason why it might be difficult to ensure that the principle of correspondence is adhered to is that the data used for the numerator and the denominator may come from different sources. For example, at an insurance company, the deaths data might be provided by the claims department whereas the exposed to risk data might come from the customer services department. These two departments might use different computer systems, different databases and different conventions, such as the age definition adopted (iil), Formuta We have been given information about the population numbers. So we can let P,(0) and P,(1) denote the number of lives aged x last birthday at the start and end of the calendar year we are considering We are also given information about the number of deaths. So we can let 0, denote the number of deaths aged x nearest birthday that occurred during the calendar year we are considering. © IFE: 2021 Examinations Page 37The estimate of v,,7 will be calculated as: where E¢ is the central exposed to risk corresponding to 0, . We can use the census method, which tells us that, for a one-year study: Fi EE = Jo Pc (that where P(t) is the number of lives at time t who are aged x nearest birthday (ie according to the age definition used in 0, for the deaths). If we assume that the population numbers vary linearly during the calendar year we are considering, we can approximate this integral using the trapezium rule: (0) +P.) | (0) is a count of the people who are aged x nearest birthday at the start of the year. Since we are not given this information directly, we will need to approximate it. If a person is aged x nearest birthday on a particular date, then they must either be x—1 last birthday or x last birthday. If we assume that birthdays for these lives are spread uniformly over the calendar year in question, there will be half of each type. So: P; (0) = $[F,-1(0) +P, (0)] Using similar logic: Pr) = $[ Pn +] So, using the census method, we can approximate E° as: Eo = ${5[F.10)+P.0)]+$[F110+.(0]} =F [Py-1(0) +P, (0) + P.-(1) + Py} Page 38 © IFE: 2021 ExaminationsWe also need to determine the value of the age adjustment f by considering the rate interval. Here the rate interval is the year when a life is aged x nearest birthday. This runs from exact age x-} to exact age x+4. We are assuming that the force of mortality is constant over the year. So the age to which the estimate relates will be the average age in the middle of the rate interval, which is x. So f=0-. So the final formula is: {Pxa1(0) + (0) + Pys(1) + PD} (iv) Factors for subdividing the data Factors that a government statistical office might use to subdivide the data include: * sex * marital status * nationality or ethnic group * — employment status. Also see Core Reading Paragraphs 3 and 4. © IFE: 2021 Examinations Page 39Subject CT4 April 2012 Question 2 ()_ Reasons for subdividing the data Mortality rates vary significantly between different categories of lives, eg males and females To reduce this heterogeneity, we can subdivide the data based on factors ‘such as age, sex, policy type and smoker status, to obtain groups that are more homogeneous. Insurance companies and other users of the data may require mortality rates for specific categories of lives, eg a 50-year-old male smoker with a term assurance policy. Subdividing data enables appropriate premiums to be charged for different individuals, which can help to avoid anti-selection. Many models of mortality assume that the lives involved all have the same rates of mortality. Subdividing the data makes this assumption more reasonable. (i) Problems with subdividing the data In practice it is not possible to make the subdivided groups completely homogeneous. Some of the subdivisions may contain little or no data, eg 20-year-olds with pension policies. Results for these groups will therefore be unreliable. It may not be possible to collect accurate information (eg because people do not give honest answers or it is prohibited by law to ask them) on certain characteristics that may affect mortality such as ethnic origin, results of genetic tests and sex. To ensure correspondence, it is important that, as far as possible, the same subdivisions are used for recording deaths and exposure. This could be a problem if, for example, somebody gives up smoking during the study. Page 40 © IFE: 2021 ExaminationsSubject CT4 April 2012 Question 9 (i) Factors by which life insurance mortality statistics are often subdivided Factors include: * sex * policy type * duration (ie how long ago the policyholder took out their policy) * ‘first class’ versus ‘impaired’ lives (ie whether the policyholder has any ‘significant known health issues). Also see Core Reading Paragraphs 3 and 4. (ii) Central death rates for calendar year 2009 The central rate of mortality for the rate interval labelled x is estimated as 0,/E¢ , where 0, denotes the number of deaths during the investigation of lives aged x last birthday. The period of investigation is one year. So: ES = [Px(thot where P, (t) is the number of policyholders aged x last birthday at time t and time is measured in years from 1 January 2009. If we assume that the population numbers varied linearly over the calendar year, then using the census method: Ef = 2[P.(0)+P,(0] For Gasperton, we have: 6, = 25 }(2,000 + 2,100) = 2,050 é 0, 25 = = -00122 © Gem = ee ~ 2.050 © IFE: 2021 Examinations Page 41Similarly, for Great Hawking: 6,=21 Ef =4(,770+1,674) = 1722 7 ~% 2A > Menon = 5 = Typ 700122 So the estimated central rate of mortality is 0.0122 for each town. (ii) Central death rates for smokers and non-smokers Let meso and m'iS., be the central mortality rates for smokers and non-smokers in Gasperton respectively. 50% of the policyholders in Gasperton are smokers and 50% are non-smokers. So: O.5 Gasp + 0.5m Gagp = 0.0112 But m&asp = 1.4m{¥§,, . assuming the same relationship as applies nationally. So: 0.5x1.4m{Sq, + 0.5m eg, = 0.0112 and hence: 0.0112 m&sp = 72 70.0106 Masp = 1.4 0.0106 = 0.0142 Similarly: 0.2 MS awk +0.8 Say = 0.0112 and again assuming the same relationship as applies nationally, eres MEHawk = 14M GHawk - So: 0.21.4 mSauy + 0.87 S 44 = 0.0112 Page 42 © IFE: 2021 ExaminationsHence: ne 0.0112 IM GHowk =~ 08, MB rawx = 1-4 «0.0113 = 0.0158 = 0.0113 (iv) Comment The insurer currently charges the same premiums for both towns. However, our crude estimates indicate that the underlying rates for both smokers and non-smokers are higher for Great Hawking. This would suggest that the company might be under-charging policyholders in Great Hawking and over-charging them in Gasperton. If the company does not differentiate on the basis of geographical area in its prices, it may lose business in Gasperton to a rival company that does differentiate. Conversely, in Great Hawking it may attract new business from rival companies, but will be under-charging and hence may risk its life assurance fund becoming insolvent. Our calculations also assume that the national differential of 40% in mortality rates for smokers applies in both Gasperton and Great Hawking, whereas the differentials may actually be different. The insurer currently charges 40% higher premiums for smokers. However, the theoretical premiums for a 25-year term assurance policy are not directly proportional to the mortality rate for a single age. So a 40% higher mortality rate does not imply a 40% higher premium rate. (This is because the calculation of the premium is affected by other factors such as interest rates and expenses.) The data we have been given only allows us to estimate the crude rates for a single age, whereas the premium calculation will depend on the mortality rates over a 25-year age range. © IFE: 2021 Examinations Page 43Subject CT4 September 2012 Question 3 () The principle of correspondence as it applies to mortality rates The principle of correspondence states that a particular day should be included in the exposed to risk for an individual life if and only if, were the Person to die on that day, they would be included in the corresponding death count. (i) Force of mortality at age 50 The estimate of the force of mortality « will be calculated as: - _&, Aaa “Ee where EF is the central exposed to risk corresponding to 0, . We have been given information about the population numbers. So we can let P,(t) denote the number of lives aged x last birthday at time ¢ . We are also given information about the number of deaths. So we can let @,(t) denote the number of deaths aged x next birthday in calendar year and P;(t) denote the number of lives aged x next birthday at time t . So: Ph(t)=Pya(t) We can use the census method, which tells us that, for a two-year study: 25 EE = fp Pa (tat We can rewrite this as: c _ (2p EE = J Pr(tat = [pPiitat +f! SPL nat +f? Prat Page 44 © IFE: 2021 ExaminationsUsing the trapezium rule on each integral, we have: Ef = F[PO+ Pico ]+ 3x 5[PL(D+ PLC.5)] +3 x S[PL (1.5) +PL(2)] Here we are assuming that the population varies linearly between census dates. Simplifying and using the fact that P; (t) = P,,(t), this becomes: 1 . 1 1 Ef = 4 PLO)+ SPL()+ PLS) + LPL) 1 3 1 1 = PPO SPD 4 BP t-6) + LP 2) From the data in the question, we get: gq (2009) + Ago (2010) 1 3 1 1 7 Paal0) +3 Paal +3 Paalt-5)+ 5 Pao(2) 200 +225 Asost = °F 3 7 7 5% 2000 + 5 2,100 + 52,300 +2 «2,500 425 =—— = 0.097701 4,350 (iii) Exact age to which this relates The force of mortality 4 corresponds to the age in the middle of the rate interval. 0, denotes the number of deaths aged x next birthday. So, any life dying must be between the ages of x-1 and x and the rate interval is [x-1,x). So the age to which our estimate of the force of mortality 4 applies is 49.5. © IFE: 2021 Examinations Page 45Subject CT4 April 2013 Question 5 () Estimate the death rate at age 65 Let P;(t) denote the number of inhabitants aged x nearest birthday at time t , measured in years from the President's birthday in 2006. The central exposed to risks for age x for each of the two periods are then: E(t) = Pat and £¢(2)= Je Prat If we now assume that the population numbers vary linearly between the census dates, we can apply the trapezium rule to approximate the integrals: E5() =3x1[PL(0)+P/(3)|_ and Ef(2)~ 1{P,(3)+P/(4)} Let P,(t) denote the number of inhabitants aged x last birthday at time f , measured in years from the President's birthday in 2006. If we assume that birthdays are distributed uniformly over each calendar year, then, of the P;(t) inhabitants aged x nearest birthday, half will be aged x-1 last birthday and half will be aged x last birthday. So: Pet) = 4 {Py s(t) +P x(O)} The census formulae then become: ES(1) = 3x34 (Py1(0) +P,(0)) + 4(Py-1(3)+F(3))} and: E502) = 1{2(P4(3)+P,(3)) +2 (Pyn(4) + Pe(A))} Page 46 © IFE: 2021 ExaminationsSubstituting the numerical values (expressed in thousands) when x = 65 gives: E(t) = 3x 3{4(Poa(0)-+Po5(0)) +4(Pea(3) + Pos(3))} =3x4{4(300+290) +4(320+310)} =3x4(295+315} =915 Similarly: E§5(2) ~ {4 (Poq(B) + Pos(3)) +3 (Poa(4) + Pag(4))} = }{4(320+310) + $(350+330)} 1 =3 {315 + 340} = 327.5 The corresponding estimates of the force of mortality are then obtained by dividing the observed numbers of deaths (again expressed in thousands) by the corresponding exposed to risk figures: figg(t) = 205). 3 _ Fea = By” ag ~ 0.00828 and: 9es(2) _ Hes(2) = Eés(2) = aye = 0.00905 This assumes that the force of mortality is constant over each year of age. (ii) Age to which the estimates apply These estimates have been derived for the rate interval labelled as 65, which is the period when the lives were aged 65 nearest birthday. This runs from exact age 64% to exact age 65%. In the middle of this period the lives were aged exactly 65. So the estimate of the force of mortality applies to exact age 65. © IFE: 2021 Examinations Page 47Subject CT4 September 2013 Question 1 (Why subdivide mortality data In human populations mortality risk varies between individuals for many reasons. However, mortality models assume that we are dealing with groups of people who have the same mortality characteristics (‘identical lives’). We can subdivide the data in order to try to achieve these homogeneous groups. This will reduce much of the heterogeneity present. (i) One problem One problem with subdividing data is that some of the subgroups may be very small, containing only a few individuals. Estimates of mortality rates derived from the small groups will be unreliable, as it will be difficult to pin down the true underlying rates with any certainty. The other main problem is incomplete data. You can discuss this instead here. (iii) Factors for subdividing the data Factors that are commonly used to subdivide mortality data include: = sex ° age * smoker / non-smoker ‘* occupation / employment status. Also see Core Reading Paragraphs 3 and 4. Page 48 © IFE: 2021 Examinations10 Subject CT4 April 2014 Question 2 () Reasons for subdividing the data Mortality rates vary significantly between different categories of lives, eg males and females. To reduce this heterogeneity, we can subdivide the data based on factors believed to have a significant effect on mortality, such as age, sex, policy type and smoker status, to obtain groups that are more homogeneous. Insurance companies and other users of the data may require mortality rates for specific categories of lives, eg a 50-year-old male smoker with a term assurance policy. Subdividing data enables appropriate premiums to be charged for different individuals, which can help to avoid anti-selection. Many models of mortality assume that the lives involved all have the same fates of mortality. Subdividing the data makes this assumption more reasonable. (ii) Four factors Apart from age and sex, mortality statistics are often subdivided by: * smoker status * policy type (for an insurance company) * known medical conditions * location / postcode. Also see Core Reading Paragraphs 3 and 4. © IFE: 2021 Examinations Page 49ah Subject CT4 April 2014 Question 4 (Principle of correspondence The principle of correspondence states that a life should be included in the exposed to risk at age x at any given time if and only if, were they to die at that time, they would be counted as a death at age x . (i) Estimate the death rate at age 45 last birthday To estimate the death rate at age 45 last birthday for the calendar year 2012 based on the two countries combined, we need to divide the total number of deaths aged 45 last birthday by the total central exposed to risk at age 45 last birthday. The total number of deaths aged 45 last birthday is the same as the total number of deaths aged 46 next birthday, which we are told is 4,800 for calendar year 2012. Country A The central exposed to risk at age 45 last birthday for Country A for the calendar year 2012 is: Eis(A) = [P&(Oat where PA(t) denotes the number of lives in Country A aged 45 last birthday t years after 1 January 2012. This corresponds to the shaded area in the graph below. To approximate this area, we can divide the shaded area into two trapeziums: * one covering the period from 1 Jan 12 to 1 Feb 12, which is of length 1 month * one covering the period from 1 Feb 12 to 1 Jan 13, which is of length 11 months. Eis(A)= (SPA (bat + ft Pe Oat Page 50 © IFE: 2021 ExaminationsCountry A 4 Jan 13? 1 Jan 12? \y385,000 ere 374,000 1Feb 11 1 Feb 12 1 Feb 13 t=0 t=1 Assuming that the population numbers vary linearly between the census dates, we can approximate these integrals using the trapezium rule to get: ES5(A) = 5 5x 3(Pub(O)+ Pas (s)) +84 (Péb (4 PACD) We can find the value of Pg5 (4) , ie at 1 Feb 12, directly from the table: PAB (75) = 381,000 To find the values of Pf (0) and Pés (1) , ie the January figures, we need to interpolate between the February figures: P45 (0) = = x 374,000 + 35 x 381,000 = 380,417 PASC) ~ 3b x 381,000 + 43 x 385,000 = 384,667 We then have: EG, (A) = =z (380,417 + 381,000) + 3 (381,000 + 384, 667) = 382,656 © IFE: 2021 Examinations Page 51Country B To match up the age definition for Country B, we can first find approximations for the population numbers based on age 45 last birthday, assuming that birthdays are uniform over the calendar year. Since a person aged 45 last birthday must be either 45 nearest birthday (if they are in the range (45,45%) ) or 46 nearest birthday (if they are in the range (45%,46)), we can average the rows for 45 and 46 to get the required approximations for age 45 last birthday: 1 Aug 2011 (374,000 + 364,000) = 364,000 1 Aug 2012: (384,000 + 372,000) = 376,500 1 Aug 2013: (385,000 + 375,000) = 380,000 The central exposed to risk at age 45 last birthday for Country B for the calendar year 2012 is: Fi5(6) = [JPR (at where P(t) denotes the number of lives in Country B aged 45 last birthday t years after 1 January 2012. This corresponds to the shaded area in the graph below. Country B (using age last birthday) 1 Jan 13? XN 380,000 1 Jan 12? 376,500 NX 364,000 1 Aug 11 1 Aug 12 1 Aug 13, t=0 t=1 Page 52 © IFE: 2021 ExaminationsAgain, assuming that the population numbers vary linearly between the census dates, we can approximate this as: Eis(B) = [PRIN at + ft PRI = 12% 2 (PIO) +P) + fo 3 (Pl (2) + PCD) We calculated the value of P%(Z) , ie at 1 August 2012, above: P& (4) = 376,500 To find the values of P/(0) and P%&(1) , ie the January figures, we need to interpolate between the August figures: PB (0) = J x 364,000 + 5 x 376,500 = 369,208 PE (1) = J x 376,500 + § x 380,000 = 377,958 We then have: E%s(B) = 4 $(369,208 + 376,500) + 5; x 4(376,500 + 377,958) = 374,677 So the estimate of the death rate for age 45 last birthday is: 4,800 ———$—— = 0.006338 382,656 + 374,677 (ii), Age to which the estimate relates We have estimated a mortality rate for the rate interval when lives were aged 45 last birthday (assuming that the force of mortality is constant over this period). This period runs from exact age 45 to exact age 46. In the middle of this period the lives were aged 45%. So the estimate applies to age 45%. © IFE: 2021 Examinations Page 5312 Subject CT4 September 2014 Question 3 () Census approximation Census approximation is a method of estimating the central exposed to risk for a population based only on population counts made at certain key dates. If P,(t) denotes the number in the population classified as aged x at time t , the central exposed to risk between censuses taken at times t; and th is equal to Ef = |? P, (that If we assume that the population number varies linearly over this period, this can be approximated using the trapezium rule to give: EE = [2 Py(t)at = (bt) « A Pelt) + P(t)] (i) Calculation of the central exposed to risk We need to find the central exposed to risk for age 55 last birthday for the calendar year 2012, ie for the period from 1 January 2012 to 31 December 2012 (which we can assume is the same as 1 January 2013). Company A Alife aged 55 last birthday must be either 55 or 56 nearest birthday. So, if we assume that birthdays are uniform over the calendar year, we can approximate the numbers aged 55 last birthday by averaging the numbers age 55 and 56 nearest birthday given in the data. So the number aged 55 last birthday on 1 January 2012 was approximately: $13,390 + 3,020) = 3,205, and on 1 January 2013: (3,100 + 2,950) = 3,025 We can use the trapezium approximation to find the exposed to risk: E£5(A) = 2(3,205 + 3,025) = 3,115, Page 54 © IFE: 2021 Examinations