THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY

ANSELM FLEISCHMANN, JONAS HIRZ

BELTIOS GmbH
Lehargasse 1, 1060 Vienna, Austria

Abstract. This paper provides a rigorous mathematical formulation of the

main components within the building block approach (BBA) and the variable
fee approach (VFA) in the IFRS17 standard as released in May 2017. The given
actuarial model provides formulas for a precise depiction of the contractual
service margin (CSM). A straightforward derivation of the insurance result is
given. Quantitative aspects for the applicability of the VFA are introduced.
Provided model extensions include the derivation of insurance revenue. De-
mystifying the complex standard, the paper illuminates critical passages, gives
rise to potential actuarial approximations tackling undue efforts and provides
a basis for implementation.

Contents

1. Introduction 2
2. The Actuarial Model 2
2.1. Actuarial Conventions for the Model 3
2.2. Notation 3
2.3. Fair Value of Assets 4
2.4. Present Value of (Future) Cash Flows 4
2.5. Risk-Adjustment for Non-Financial Risk 7
3. CSM 9
3.1. CSM at Initial Recognition 9
3.2. Subsequent Measurement of the CSM 10
4. Insurance Contract Liability and Insurance Result 14

E-mail address: [email protected], [email protected].

Date: July 29, 2018.
Key words and phrases. IFRS17, specification of IFRS 17, formulas, actuarial model, BBA,
GMM, VFA, underlying, universal underlying, CSM, insurance result, revenue.
The authors want to thank the working groups for IFRS within the Austrian actuarial association
AVÖ as well as within the German actuarial association DAV for all the intense and fruitful
discussions giving insight into the highly complex IFRS17 standard.
1

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
2 A. FLEISCHMANN, J. HIRZ

5. Insurance Revenue 15
6. Conclusion and Outlook 16
7. Proofs 16
References 17

1. Introduction

The International Accounting Standards Board issued the IFRS17 standard for
insurance contracts in May 2017. With an effective date of January 1, 2021, it is set
to replace the IFRS 4 standard. Whilst being an accounting standard, IFRS 17 will
require extensive actuarial and IT expertise to meet the requirements set out in the
standard.
Since issuance of the standard, numerous opinions and statements have been circu-
lating among the industry. Within this current state of sciolism, this paper primarily
targets the perplexed actuary exposed to the field of IFRS17.
This goal shall be met by giving a short and concise depiction1 of the standard
starting with an actuarial model of the building block approach (short: BBA, also
referred to as the general measurement model or short GMM) and the variable
fee approach (short: VFA) in section 2. The premium allocation approach (short:
PAA) is essentially a simplification to the building block approach and, therefore,
not covered in this paper.
A rigorous notation, familiar to actuaries, is introduced. It allows a precise formu-
lation of reconciliations from one accounting period to another using the concepts
of random variables, filtrations and conditional expectations (see, for example,
Williams [5]). Another mathematical approach towards IFRS 17 with illustrative
examples can be found in the Master Thesis of Widing and Jansson [4]. Selected
extensions to the actuarial model are discussed in section ??.

2. The Actuarial Model

The aim of this section is to set up an actuarial model which is able to quantify the
insurance result within the BBA and the VFA. Note that the insurance result is
derived from basic accounting principles as the change of assets minus the change
in liabilities over time.
Moreover, all quantities which are required to compute insurance revenue as well
as insurance expenses including their contribution to finance and insurance service
expenses are mathematically specified in this section.
Insurance contract liabilities need to be recognised for each group of insurance
contracts according the requirements as, for example, specified in IFRS 17.32/40.
In practical terms, this may require computing results on a more aggregate level
and allocating these results to the level of groups of insurance contracts (GICs).
1
Although the provided results were derived with the utmost actuarial rigour, they are subject
to the authors’ interpretation of the standard. Other interpretations may lead to different results.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 3

However, there may be limitations to this approach, in which case exceptions apply,
see especially IFRS 17.B71 and BC138.

2.1. Actuarial Conventions for the Model. In order to boil down formulas to
the essence of the standard, the following refining assumptions were made:
(a) A portfolio with a single group of insurance contracts is considered.
(b) Claims payments occur at the same time as insured events. Within the standard
this translates into insurance contract liabilities equaling insurance contract
liabilities for remaining coverage (LRC), i.e. insurance contract liabilities for
incurred claims (LIC) are zero.
(c) There are no experience adjustments meaning, that observed cash flows come
about as expected. In particular, experience adjustments arising from premiums
that relate to future service do not occur (see IFRS 17.B96(a)).
(d) There are no investment components.
(e) Invoking IFRS 17.B77, all cash flows vary based on the returns of underlying
items within the VFA, i.e. there are no cash flows to the policyholder that do
not vary with the returns of the underlying items. Henceforth, no separate
discounting factor (see IFRS 17.B74) for cash flows that do not vary with the
returns of the underlying items needs to be introduced.
(f) The effect of any currency exchange difference, which adjust the the contractual
service margin, are equal to zero.

2.2. Notation. Embedded in a continuous-time setting, consider discrete account-

ing dates t ∈ {0, 1, . . . , T } with no cash flows beyond T ∈ N, constituting the
contract boundary. Time t = 0 indicates the time of initial recognition.
Definition 2.1 (Accounting intervals and notation). Accounting intervals are
assumed to be left-closed and, for notational simplicity, denoted by the left endpoint2,
i.e. period t constitutes cash flows in [t, t + 1).

Sigma-algebras At and Tt provide the financial (referring to the asset filtration)

and the insurance technical (referring to the technical filtration as in Wüthrich and
Merz [6]) information, respectively, available at time t ∈ [0, T ]. For all s, t ∈ [0, T ]
with s ≤ t it holds that As ⊂ At and Ts ⊂ Tt . The joint information is given by the
smallest sigma-algebra Ft = σ(At , Tt ) generated by At and Tt 3.
Definition 2.2 (Discount factors). Discount factors for cash flows due at time
u ≥ 0 measured in monetary units at time 1 ≤ t < u, with interest rate term
structure matching asset returns backing those same cash flows, are denoted by
zt,u > 0. The value of zt,u is known at time u, i.e. zt,u is Au -measurable. For
simplicity, we consider the convention zt,t = 1.

As a notational convention, cash flow quantities are denoted with lower case letters
while balance sheet quantities are denoted with capital letters.

2 For notational purposes, quantities which generate a cash flow in period [t, t + 1) are denoted
just using subscript t.
3 In accordance with the definition of accounting intervals, all corresponding filtrations are
assumed to be right-continuous.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
4 A. FLEISCHMANN, J. HIRZ

2.3. Fair Value of Assets. Typically, an entity holds assets and undertakes in-
vestment activities in order to meet its obligation to make claims payments and in
order to meet capital requirements. In this paper, it is assumed that the value of
these assets can be measured at fair value4. The notion of assets – and fair values
thereof – is required to derive the insurance result from general accounting principles
as the change in assets minus the change in liabilities.
In this paper, the introduced assets are solely those used to back the insurance
contract liabilities of the group of insurance contracts in consideration5. Within the
VFA, it is further assumed that the fair value of assets is equal to the fair value of
the pool of underlying items, see section 3.2.3.
Definition 2.3 (Fair value of assets). The fair value of assets at time t is denoted
by Ut and known at t, i.e. Ft -measurable. Given period t > 0, the expected return
on assets is defined by  
U 1
wt = E U − 1 Ft Ut
zt,t+1
where the corresponding discount factor6 zt−1,u
U
is Fu− -measurable for every u > t−1.
Furthermore,
T
X −1
U
Wt = E[zt,v |Ft ]wvU
v=t
denotes the present value of returns at t.

Once defined, the change in the fair value of assets over time can be decomposed.
Lemma 2.4 (Analysis of movement). The total change in the fair value of (under-
lying) assets is given by
U
∆t U = Ut − Ut−1 = wt−1 + cU
t−1 + ∇t F
U

with the valuation adjustment for an instantaneous change from expected to observed
fair values being given by
∇t F U = Ut − E[Ut |Ft−1 ]
and with the residual change due to (net) observed cash flows being given by
cU U
t−1 = E[Ut |Ft−1 ] − Ut−1 − wt−1 .

2.4. Present Value of (Future) Cash Flows.

Definition 2.5 (Present value of cash flows). Given cash flows cu due in period
u ≥ t, which typically include insurance technical cash flows such as premiums,
costs and claims, the present value of cash flows is measured prospectively by
T
X
FtBE = E[zt,u cu |Ft ] (2.6)
u=t

where cu is known at known at the end of period u, i.e. is F(u+1)− -measurable7.

4 See IFRS13.
5 In applications, an attribution of fair values of assets to corresponding groups of insurance
contracts may be required, as assets are typically not labelled.
6 Following IFRS 17.36(b) it may reasonably be assumed that discount factors for (underlying)
U
assets backing liabilities zt−1,u approximately match discount rates for cash flows zt−1,u .
7 Note that F S
(u+1)− = t<u+1 Ft denotes information just before time u + 1.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 5

Remark 2.7. Various passages in the standard require splits of cash flows cu into
component parts, cf. IFRS 17.B65. For sake of readability, this paper avoids such a
split unless stated otherwise. It is common to split cu into premiums and service
cash flows. The latter may contain cash flows related to claims, costs, or investment
components and may require two time indices – date incurred and date paid – giving
rise to liabilities for incurred claims8. Note that the contractual service margin only
relates to the liability for remaining coverage, as outlined in IFRS 17.B97(b).
Remark 2.8. The definition of FtBE in (2.6) allows discount rates zt,u and cash flows
cu to be stochastically dependent. Henceforth, in the spirit of IFRS 17.B39/B77,
the introduced model includes stochastic settings which account for options and
guarantees embedded in the contract. IFRS17 does not give a formula how such
embedded guarantees and options should be calculated. In particular, FtBE appears
conceptually similar to the best estimate liability in Solvency II9.
Remark 2.9. Since IFRS 17 specifies fulfilment cash flows as liabilities, cu > 0
indicates outgoing cash flows, e.g. claims payments, and cu < 0 indicates incoming
cash flows, e.g. premiums.
BE
Definition 2.10 (Unwinding). Given period t > 0, unwinding wt−1 denotes the
BE
expected accretion of interest of Ft moving from t − 1 to t,
T
X
BE
wt−1 = E[(zt,u − zt−1,u )cu |Ft−1 ] .
u=t

Remark 2.11. Assuming that zt−1,u = zt−1,t zt,u for all u ≥ t and that zt−1,t is
At−1 -measurable, then
 1 
BE BE


wt−1 = − 1 Ft−1 − E[ct−1 |Ft−1 ] ,
zt−1,t
which states that unwinding equals the accretion of interest on the present value of
cash flows after the release of the expected cash flow in period t − 1.
Definition 2.12 (Valuation adjustment). Given period t > 0, ∇t F BE denotes
the valuation adjustment in FtBE attributed to changes of future cash flows via
information gain with
∇t F BE = ∇t ABE + ∇t T BE
where ∇t ABE denotes the valuation adjustment attributed to changes caused by
financial information gain
T
X
BE


∇t A = E[zt,u cu |σ(At , Tt )] − E[zt,u cu |σ(At−1 , Tt )] + ψ(A0,t−1 , Tt−1,t )
u=t

with correction term ψ(A0,t−1 , Tt−1,t ), as discussed in the next section, as well as
where ∇t T BE denotes the valuation adjustment attributed to changes caused by

8 By convention in this paper, the liability of incurred claims equals zero.

9 DIRECTIVE 2009/138/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
of 25 November 2009 on the taking-up and pursuit of the business of Insurance and Reinsurance
(Solvency II)

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
6 A. FLEISCHMANN, J. HIRZ

insurance technical information gain10

T
X
BE


∇t T = E[zt,u cu |σ(At−1 , Tt )] − E[zt,u cu |σ(At−1 , Tt−1 )]
u=t
− ψ(A0,t−1 , Tt−1,t ) .

Remark 2.13. The valuation adjustment ∇t F BE denotes the instantaneous change

in present value of (future) cash flows. In particular, it accounts for an unexpected
change in coverage units and corresponding future cash flows for the following
reasons:

(a) Due to variation in lapse or mortality rates more or less contracts than previously
expected may constitute a group.
(b) Additional contracts are added to a group (see IFRS 17.28) where it is assumed
that these newly added contracts do not exhibit cash flows prior to t, i.e. ev-
erything that happens between t − 1 and t (which is known at t) is treated in
analogy to an experience variance as required by IFRS 17.BC233.
(c) Contracts are derecognised from a group (see IFRS 17.72–77).
Lemma 2.14 (Analysis of movement). The change in present value of future cash
flows from time t − 1 to t is given by
∆t F BE = FtBE − Ft−1
BE BE
= wt−1 − cBE
t−1 + ∇t A
BE
+ ∇t T BE

where cBE
t−1 = E[ct−1 |Ft−1 ] for the (nominal) cash flows in period at t − 1 expected
the beginning of this period.

According to the lemma above, the change between t − 1 and t of liabilities is defined
in a prospective way and can be expressed by the following components:

(a) Unwinding, i.e. the accretion of interest as expected at t − 1

(b) Release of expected cash flows due in period t − 1
(c) Valuation adjustments for future cash flows (i.e. ≥ t)

2.4.1. Adjusting Technical Changes to the Financial Framework A0 . For updating

the contractual service margin using the BBA, IFRS 17 requires the basic change in
technical assumptions without the correction term to be further split down into a
component measured according to the financial framework of A0 . The remaining
difference leads to the correction term
T 
X

ψ(A0,t−1 , Tt−1,t ) = E[zt,u cu |σ(At−1 , Tt )] − E[zt,u cu |σ(A0 , Tt )]
u=t


− E[zt,u cu |σ(At−1 , Tt−1 )] − E[zt,u cu |σ(A0 , Tt−1 )]

10 Technical changes are observed in technical frameworks T

t−1 and Tt but have to be expressed
in terms of financial framework at initial recognition A0 , see IFRS 17.B72(c). Note that the order,
in which the insurance technical filtration and the financial filtration are updated in this paper, is
arbitrary as the standard does not specify this issue. Similar problems arise in accounting when it
comes to simultaneous quantity and price deviations.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 7

see IFRS 17.B72(c)11, that transposes the change in future technical assumptions
of the present value of cash flows made between t − 1 and t to the financial (asset)
framework A0 .
In the BBA, the contractual service margin is maintained based on the financial
framework A0 whereas present values of cash flows are maintained based on the
actual financial framework At−1 at beginning of period and At at end of period.
Since changes in future cash flows affect both liabilities these changes need to be
transposed between A0 and At−1 for reconciliation.
This is where actuaries tend to get perplexed, because:

(a) The requirement to maintain one part of the insurance contract liability (the
CSM) based on the financial framework A0 and all other parts based on the
current financial frameworks At−1 at beginning of period and At at end of
period, requires the computation of the correction term ψ(A0,t−1 , Tt−1,t ) for
every reconciliation12. Such an approach is regarded genuinely impractical.
(b) Given the limitations to the accuracy of these terms13, it is by no means clear
to actuaries why the accumulation of inaccuracies should lead to meaningful
information for readers of financial statements rather than generate spurious
accuracy.
(c) This approach may lead to the weird case where the loss carried forward is
strictly greater than zero while the liability for remaining coverage is strictly
less than zero, i.e. is an asset. A proposal how to treat this exception is outlined
later.

Remark 2.15. Within the VFA, ψ(A0,t−1 , Tt−1,t ) need not be computed, as it is not
required to use discount rates determined for derivation of the CSM at the date of
initial recognition. However, the case with the loss carried forward strictly greater
than zero and liability for remaining coverage strictly less than zero cannot be ruled
out.

2.5. Risk-Adjustment for Non-Financial Risk. In this paper, the risk adjust-
ment for non-financial risk is defined as a conditional risk measure, see IFRS 17.11914
that is kept rather general in order to include various modelling and allocation
approaches.

11 ‘to measure the changes to the contractual service margin applying paragraph B96(a)–(c)
for insurance contracts without direct participation features–discount rates applying paragraph 36
determined on initial recognition.’
12 and literally also for every group of insurance contracts.
13 In real-world cash flow models of portfolios of insurance contracts there are several factors
limiting the computational accuracy of analysis-of-movement calculations, such as unexplained
movements, partially accounted changes in perimeter, etc. Furthermore, note that the order of
computation for the determination of changes in financial assumptions and changes in technical
assumption as well as often the classification of experience variances as technical versus financial
are regarded arbitrary.
14 ‘An entity shall disclose the confidence level used to determine the risk adjustment for
non-financial risk. If the entity uses a technique other than the confidence level technique for
determining the risk adjustment for non-financial risk, it shall disclose the technique used and the
confidence level corresponding to the results of that technique.’

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
8 A. FLEISCHMANN, J. HIRZ

Definition 2.16 (Risk adjustment for non-financial risk). Given period t ≥ 0, let
bu denote the T(u+1)− -measurable components for all u ≥ t with calculation basis
T
X
bt,T = bu .
u=t

Then, the risk adjustment for non-financial risk15 is given by

Rt = ρ[bt,T | Tt ] .
Remark 2.17. Note that since the calculation basis bt,T is measurable with respect
to the insurance technical filtration, the risk adjustment is derived solely based
on non-financial risk. In particular, the definition covers settings where the risk
adjustment is derived from other measurements, e.g. as in Solvency II, or any
allocations thereof.
Definition 2.18 (Fulfilment cash flows). According to IFRS17, fulfilment cash
flows Ft are given by Ft = FtBE + Rt .
Definition 2.19 (Release). Given period t > 0, the release of risk margin xR
t−1 due
to service provided in period t − 1 is defined as the corresponding change in risk,
measured at t − 1, i.e.
xR
t−1 = Rt−1 − ρ[bt,T | Tt−1 ] .

Remark 2.20. Let ρ[S, bt−1,T | Tt−1 ] denote the risk contribution of sub risk S to
the total calculation basis bt,T , cf. Kalkbrener [2] for the classical case16. Assuming
linearity of risk contributions, by definition we have
Rt−1 = ρ[bt−1 , bt−1,T | Tt−1 ] + ρ[bt,T , bt−1,T | Tt−1 ] .
Henceforth,
xR


t−1 = ρ[bt−1 , bt−1,T | Tt−1 ] − ρ[bt,T | Tt−1 ] − ρ[bt,T , bt−1,T | Tt−1 ] ,
i.e. the release of risk margin can be written as the contribution of component bt−1
to the total calculation basis bt−1,T minus an adjustment for a loss in diversification
due to the missing time slice t − 1.
R
Definition 2.21 (Unwinding). Given period t > 0, unwinding wt−1 denotes the
expected accretion of interest on the risk adjustment, after any release, moving from
t − 1 to t  
R 1
wt−1 = E R − 1 At−1 ρ[bt,T | Tt−1 ] ,
zt−1,t
R
where discount rates zt−1,t > 0 are At -measurable.
Remark 2.22. Note that, according to the standard, the entity may choose not to
R
include interest accretion zt−1,t on the risk adjustment for non-financial risks (see
17
IFRS 17.81 ) or comply to the rate being applicable to nominal cash flows that do
not vary based on the returns on any underlying items.
15 The notation is taken from [1, Chapter 1].
16 In this work, three axioms are proposed. one of these axioms is linearity which states that
‘the risk capital of the portfolio equals the sum of the (contributory) risk capital of its subportfolios’.
17 ‘An entity is not required to disaggregate the change in the risk adjustment for non-financial
risk between the insurance service result and insurance finance income or expenses. If an entity
does not make such a disaggregation, it shall include the entire change in the risk adjustment for
non-financial risk as part of the insurance service result.’

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 9

Definition 2.23 (Valuation adjustment). Given period t > 0, ∇t T R denotes

the valuation adjustment for the instantaneous change due to changes in (future)
technical cash flows via information gain with
 
R 1
∇t T = Rt − E R At−1 ρ[bt,T | Tt−1 ] .
zt−1,t

Remark 2.24. In analogy to ∇t T BE , new and modified business can lead to valuation
adjustments for future technical cash flows.
Lemma 2.25 (Analysis of movement). The movement of the risk adjustment for
non-financial risk from time t − 1 to t is, similarly as for FtBE , given by
R
∆t R = Rt − Rt−1 = wt−1 − xR
t−1 + ∇t T
R
.

3. CSM

The idea of the contractual service margin (shorthand CSM) is to measure unearned
profits for future insurance service. These profits will later be recognised in the
statements of financial performance when the contractual services margin is allocated
according to the services provided.
As mentioned in remark 2.7, the liability for incurred claims is not related to future
service (i.e. coverage) and so does not affect the CSM. As, by convention, the
liability for incurred claims equals zero, it is not necessary to introduce an additional
notation with superscript ‘RC’. However, in order to avoid errors when applying
the stated formulas, the reader should be aware that all quantities in this section
solely relate to remaining coverage.

3.1. CSM at Initial Recognition. The CSM at initial recognition18 is, according
to IFRS 17.38, defined identically for the BBA and the VFA as follows:
Definition 3.1 (CSM and loss carried forward at initial recognition). At initial
recognition t = 0, the CSM is given by
LCSM

0 = − min F0 , 0
where F0 denotes fulfilment cash flows for remaining coverage19 at t = 0, while the
loss carried forward is given by
LLoss

0 = max F0 , 0 .

The CSM and the loss carried forward serve as corresponding vessels where either
one is positive while the other equals zero at any instant of time. The loss carried
forward needs to be transformed to a loss component of the liability for remaining
coverage.

18 Initial recognition (of a group of insurance contracts) also occurs at modification date or
with business transfer/business combinations.
19 At initial recognition, it is reasonable to assume that there is no liability for incurred claims.
However, there are possible cases with incurred claims being strictly greater than zero at initial
recognition (e.g. business combinations).

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
10 A. FLEISCHMANN, J. HIRZ

3.2. Subsequent Measurement of the CSM. In contrast to the present value

of future cash flows and the risk-adjustment for non-financial risk, the CSM is
not defined in a prospective way at subsequent measurement dates. Instead, it is
updated retrospectively and includes prospective valuation adjustments. The same
holds for the loss carried forward.
Due to unfavourable changes in future cash flows the profitability of a group of
insurance contracts decreases and may give rise to a loss, i.e. the GIC becomes
onerous in terms of the standard. For an onerous group of insurance contracts,
an entity shall establish a loss component of the liability of remaining coverage,
as outlined in IFRS 17.49–52. This loss component is closely related to the loss
carried forward LCSMt and ‘determines the amounts that are presented in profit or
loss as reversals of losses on onerous groups and are consequently excluded from the
determination of insurance revenue’. It shall be noted that the loss component is not
an additional liability, but a part of (i.e. a component) the liability for remaining
coverage.

3.2.1. Loss component. In order to cover all possible cases, the following definition
makes a distinction between weakly and strictly onerous groups of contracts.
Definition 3.2 (Strictly and weakly onerous). A group of insurance contracts is
called strictly onerous at t ≥ 0 whenever Ft > 0 and LLoss
t > 0. A group of insurance
CSM
contracts is called weakly onerous whenever Lt = 0.
Remark 3.3. The notion of weakly onerous groups includes the case where LLoss
t >0
hence LCSM
t = 0 and the liability for remaining coverage Lt equaling F t < 0 being
an asset.
Definition 3.4 (Loss component). The proportion of the loss component in relation
to the fulfilment cash flows for remaining coverage20 λt is given by
( n Loss o
L
min Ft t , 1 if the group is strictly onerous
λt =
0 else .
Hence, the loss component LC
t is given by

LC
t = λt Ft .

Remark 3.5. Note that λt > 0 describes a situation where there is a positive loss
component (see the definition below) in which case the CSM is zero. Furthermore,
the case where Ft ≤ 0 and LLoss
t > 0 is treated in the same way as the degenerate
Loss
case Ft = 0 and Lt = 0.

In order to define the systematic depreciation of the loss component, as required by

IFRS17.50(a) and 52, the basic idea is that the loss component of the liability of
remaining coverage evolves during the period before any new valuation adjustments
pop up at the end of the period – in principal – in the same way as for the liability for
remaining coverage. Henceforth, the systematic depreciation of the loss component
is defined as follows:
20 The definitions of λ relies on LLoss and F where the first quantity is defined retrospectively
t t t
while the latter quantity is defined prospectively.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 11

Definition 3.6 (Systematic depreciation factor of the loss component). At time

t ≥ 1, the systematic21 depreciation factor dt−1
C
of the loss component is defined by
( n o
C min 1 − FFt−1t
,1 Ft−1 > 0
dt−1 =
0 else .
C
Remark 3.7. The specification allows that dt−1 < 0, leading to an increase of the
loss component if also the liability for remaining coverage increases.
Remark 3.8. The systematic depreciation of the loss component must not cause a
reversal of losses (see IFRS 17.49) implying that a reversal of losses can only occur
C
due to favourable valuation adjustments, hence dt−1 ≤ 1.

3.2.2. The CSM Formula within the BBA.

CSM
Definition 3.9 (Unwinding). Given period t > 0, unwinding wt−1 denotes the
expected accretion of interest on the CSM, moving from t − 1 to t, measured at
discount rates determined at the date of initial recognition, i.e.
h 1 i
CSM
wt−1 = E − 1 A0 LCSMt−1 .
zt−1,t

For notational purposes, the following auxiliary variables are introduced.

Definition 3.10 (Auxiliary variable). Given period t > 0, define
Gt = LCSM CSM
t−1 + wt−1
as well as
C
Ht = λt−1 Ft−1 (1 − dt−1 ) + ∇t T BE + ∇t T R
Then, the CSM remaining at the end of the period before any allocation for services
provided is defined by
LCSM
t = max{Gt − Ht , 0} .
Definition 3.11 (Allocation of the CSM for services provided22). Defining coverage
(service) units su ≥ 0 for u ≥ t − 1, which are F(u+1)− -measurable23, we set
st−1
τt−1 = PT ∈ [0, 1]
st−1 + u=t E[su |Ft −]
PT
whenever st−1 + u=t E[su |Ft− ] > 0. The allocation aCSM t−1 of the contractual
service margin remaining at the end of the reporting period, which is recognised as
insurance revenue because of the transfer of services in the period, is given by
aCSM CSM
t−1 = τt−1 Lt .
Remark 3.12. Details regarding the derivation of coverage units are given in IFRS
17.B119. It shall be outlined that the standard explicitly requires the derivation of
coverage units actually provided (and not expected) in the current period t − 1.
21 The standard does not preclude other drivers from which a depreciation factor of the loss
component of the liability for remaining coverage may be derived as long as they are systematic.
22 See IFRS 17.44(e): ‘the amount recognised as insurance revenue because of the transfer of
services in the period, determined by the allocation of the contractual service margin remaining at
the end of the reporting period (before any allocation) over the current and remaining coverage
period applying paragraph B119.’
23 Note that s
t−1 denotes the observed coverage units at time t as it is Ft− -measurable.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
12 A. FLEISCHMANN, J. HIRZ

Based on the notation for the movement of prospective liabilities, the evolution of
the contractual service margin as well as its shadow, the loss carried forward, can
be defined as follows:
Definition 3.13 (CSM and loss carried forward for t > 0 within the BBA). Given
the above definitions, for period t > 0,
LCSM
t = max{Gt − Ht − aCSM
t−1 , 0}
as well as
LLoss
t = − min{Gt − Ht , 0} .
Remark 3.14. Invoking the definition of the CSM, it follows that
LCSM
t = max{Gt − Ht , 0} − aCSM CSM
t−1 = (1 − τt−1 )Lt .
Example 3.15. It is interesting to analyse the case LLoss
t−1 > 0 and Lt
CSM
> 0. In
this case, a group of contracts onerous at t − 1 becomes not onerous at t due to
favourable changes in service cash flows ≥ t, hence LCSM
t > 0, yielding aCSM
t−1 > 0.
24
This contradicts the IFRS17 guiding principle that profits arising from future cash
flows shall not be recognised in the current period.

3.2.3. The CSM Formula within the VFA. When it comes to the treatment of the
CSM25, the standard requires entities to distinguish between insurance contracts
without (see IFRS 17.44) and with (see IFRS 17.45) direct participation features.
The so-called VFA is applied to insurance contracts with direct participation features
where details are given in IFRS 17.B101–118. Most importantly, application of
the VFA is mandatory if and only if an assessment of all three criteria IFRS
17.B101(a)–(c) are positive.
Definition 3.16 (Pool of underlying items). The fair value of the pool of underlying
items at time t is given by the fair value of assets Ut .

This definition is based on the convention that there are no cash flows not varying
with the pool of underlying items. Henceforth, without loss of generality, an entity
may reasonably expect
(a) that any pool of underlying items arising from investment activities performed
by the entity is supplied by premiums26 paid by policyholders as, see IFRS
17.BC242(a),
(b) the assets backing the contractually defined pool of underlying items may be
measured at fair value,
(c) at time t = 0, immediately before initial premiums become due, the fair value
of these assets is zero.
24 See IFRS 17.BC280 and, in particular, IFRS 17.43: The contractual service margin at the
end of the reporting period represents the profit in the group of insurance contracts that has not
yet been recognised in profit or loss because it relates to the future service to be provided under
the contracts in the group. However, IFRS 17.34 does not specify that all of these future profits
are kept in the CSM.
25 In IFRS 17.BC257 it is noted that differences between GMM and VFA are limited to the
treatment of the contractual service margin.
26 These are commonly referred to as savings premiums. Hence, the present value of premiums
may be assumed as an upper bound for the fair value of any pool of underlying items at the time
of initial recognition.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 13

Definition 3.17 (Entity’s share). As required by IFRS 17.45(b), the entity’s share
of the change27 in the fair value of the pool of underlying items at time t > 0 is
Ft -measurable and denoted by δtU .
Remark 3.18. In exchange for the future service provided, a variable fee (see IFRS
17.B104) is charged by the entity, which is based on the entity’s share of the fair
value of the underlying items.

The derivation of the CSM at initial recognition, i.e. for t = 0, is identical to the
BBA. After initial recognition, there are differences in the definition of the auxiliary
variables.
Definition 3.19 (CSM and loss carried forward for t > 0 within the VFA). Within
the VFA, the CSM and the loss carried forward are obtained similarly as for the
BBA where, in contrast to definitions 3.10 and 3.13, ∇t T BE + ∇t T R is replaced by
δtU and no accretion of interest on the CSM occurs, so wt−1
CSM
= 0.

The VFA is regarded a significant improvement compared to the BBA from an

actuarial point of view with the following characteristics:
(a) The VFA allows the CSM to buffer changes in underlying items (which are
expected to be represented as assets) at the cost that technical valuation
adjustments are buffered less. In particular, this implies that asset liability
management can yield reduced variability in insurance results.
(b) In many cases, a change in the fair value of underlying items will be more
reliably measurable than ∇t T BE + ∇t T R28.
(c) IFRS 17.45 allows to determine all adjustments of the CSM as a combined
amount, unless components are required for the derivation of reconciliations.
Remark 3.20 (Conditions for VFA applicability). According to IFRS 17.B101, an
application of the VFA is mandatory if and only if an assessment of the three criteria
(a)–(c) is positive. Henceforth, the following opinions may help the reader with the
assessment of the these criteria:
(a) Qualitative: It is required that ‘the contractual terms specify that the pol-
icyholder participates in a share of a clearly identified pool of underlying
items’ where the link to the pool of underlying items has to be legally enforce-
able and not changeable by the entity in retrospect, see IFRS 17.B101 and IFRS
17.B105/106.
(b) Quantitative: It is required that ‘the entity expects to pay to the policyholder an
amount equal to a substantial29 share of the fair value returns on the underlying
items’. The present value of returns WtU can be used to verify the substantiality
condition.
27 In IFRS17, neither change nor share are explicitly specified or defined. Following a teleological
interpretation of the VFA, the authors choose the above definition. However, there may exist other
valid interpretations of the standard. It shall be noted that IFRS17 mentions both (by coincidence
or not), the entity’s share of the change as well as the change of the entity’s share.
28the authors doubt that ∇ T BE can be calculated without undue cost and effort in real-world
t
models.
29
The standard does not specify the terms expects and substantial. However, derivations based
on a set of stochastic scenarios may help to quantify decision rules.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
14 A. FLEISCHMANN, J. HIRZ

(c) Quantitative: It is required that ‘the entity expects a substantial proportion

of any change in the amounts to be paid to the policyholder to vary with the
change in fair value of the underlying items’. Within the proposed setting, a
statistical test aiming at the detection of dependence between (∆t U )t≥1 and
(∆t F )t≥1 may be conducted to verify substantiality.
• For each t ≥ 1, the correlation coefficient
E[(∆t U − E[∆t U ])(∆t F − E[∆t F ])]
ρt = p
Var(∆t U ) Var(∆t F )
can be derived for given financial and non-financial scenarios. Whenever a
substantial amount of periods shows a correlation above some threshold
α > 0, then criterion (c) is satisfied.
• Under additional assumptions, (∆t U )t≥1 and (∆t F )t≥1 can be regarded
as a correlated stochastic processes, cf. Tend, Ehrhardt and Günther [3].

4. Insurance Contract Liability and Insurance Result

Definition 4.1 (Insurance contract liability). The insurance contract liability Lt

at time t ≥ 0 is given by the sum of the present value of fulfilment cash flows and
the CSM, i.e.
Lt = FtBE + Rt + LCSM
t = Ft + LCSM
t .

Defining the insurance result xt−1 as the change in fair values of assets and liabilities
from time t − 1 to t, we get
xt−1 = ∆t U − ∆t F BE − ∆t R − ∆t LCSM . (4.2)

The insurance result under IFRS 17 for the BBA and the VFA can thus be derived
as follows:
Lemma 4.3 (Insurance result). Recalling the convention that experience variances
are assumed to be zero, i.e. cBE U
t−1 = ct−1 , for the BBA we get

xt−1 = wt−1 + ∇t F U − ∇t ABE + aCSM R

t−1 + xt−1 + `t ,

while for the VFA we get

xt−1 = wt−1 + ∇t F U − ∇t ABE + aCSM R U
t−1 + xt−1 + δt − ∇t T
BE
− ∇t T R + `t ,
U BE R CSM
where wt−1 = wt−1 − wt−1 − wt−1 − wt−1 and where `t is a correction term that
indicates the part of the change in the loss carried forward that is not buffered by
the CSM
`t = LLoss
t
C
− λt−1 Ft−1 (1 − dt−1 ).

A contested issue amongst actuaries is the question on how to explicitly define un-
derlying items for specific tariffs. While there supposedly exist numerous definitions
of the entity’s share δtU of the change in the fair value of the pool of underlying items
that are conform with the standard, a rather universal interpretation is provided in
the following corollary.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 15

Corollary 4.4 (Universal interpretation of δtU ). Defining the entity’s share δtU of
the change in the fair value of the pool of underlying items as
δtU = ∇t F BE + ∇t T R − ∇t F U + wt−1
BE R
+ wt−1 U
− wt−1
implies that, according to Lemma 4.3, the insurance result within the VFA is given
by
xt−1 = aCSM R
t−1 + xt−1 + `t . (4.5)
Henceforth, the valuation adjustment of remaining coverage associated to insurance
service (i.e. ∇t T BE within the BBA) can be evaluated as
∇t F BE − ∇t F U .
Remark 4.6. Equation (4.5) is able to unearth the impact of the VFA on the
insurance result and its fundamental difference to the BBA. All the variability of
assets and liabilities is, given the stated actuarial conventions, buffered fully by
the CSM as long as the group of insurance contracts does not become onerous. In
particular, the entity gets rewarded for proper asset liability management as the
variability in the CSM is reduced if δtU is small. The following extreme cases within
the VFA can be considered:
(a) Perfect hedge: If valuation adjustments of future cash flows are exactly replicated
by valuation adjustments of the underlying, i.e. ∇t F BE = ∇t F U , they do not
adjust the CSM30.
(b) Zero hedge: If there are no investments or if all investments are put into an asset
where corresponding valuations do not depend on discount rates (e.g. a bank
account) such that ∇t F U = 0, then the underlying does not buffer insurance
valuation adjustments and, hence, these need to be absorbed by the CSM or
contribute to a loss component.

5. Insurance Revenue

According to IFRS 17.83, insurance revenue shall depict the provision of coverage
and other services arising from the group of insurance contracts at an amount that
reflects the consideration to which the entity expects to be entitled in exchange for
those services.
It shall be noted that IFRS 17.B120–124 provide several definitions of insurance
revenue. In this paper, the definition given in IFRS 17.B124 is considered due to its
applicability within the proposed model.
Deferred acquisition costs cDAC 0 need to be considered when the insurance contract
liability is initially recognised over the contract boundary α = cDAC
0 /T . Amortisation
of deferred acquisition cash flows has to be systematic based on the passage of time.
Deferred acquisition cost need to be directly attributable to the portfolio to which
the group of insurance contracts belongs.
Note that α is reported in the statements of financial performance both as insurance
revenue and as insurance service expenses (evening each other out) based on the
passage of time. Hence, α has no impact on the insurance result.
30 In contrast to the VFA, within the BBA technical valuation adjustments of future cash flows
adjust the CSM without any buffering possibilities.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
16 A. FLEISCHMANN, J. HIRZ

As specified in 17.B120–124, the systematic depreciation of the loss component is

deducted from revenue, yielding the following formula:
Definition 5.1 (Insurance revenue). Revenue for the period t − 1 is given by
insurance BE,service

rt−1 = −aCSM
t−1 + α − (1 − λ t ) x R
t−1 + c t−1

where cBE,service
t−1 denotes service cash flows, which do not include premiums.

6. Conclusion and Outlook

In this paper, we provide a mathematical depiction of the BBA and VFA under
IFRS17. The given formulas are boiled down to the essence of the standard and are
geared to the purpose of providing a basis for implementation.
Topics within IFRS17 which are not analysed on a deeper level in this paper include
investment components, separation of cash flows into varying and not varying with
the returns of the underlying within VFA, modifications of contracts, higher levels of
aggregation, the premium allocation approach, reinsurance and other comprehensive
income.

7. Proofs

Proof of Lemma 4.3. The proof is given for the BBA. The result for the VFA follows
CSM
analogously by setting wt−1 = 0 and by replacing ∇t T BE + ∇t T R with δtU . By
(4.2), we have
xt−1 = wt−1 + ∇t F U − ∇t ABE + xR
t−1 − ∇t T
BE
− ∇t T R − ∆t LCSM . (7.1)
Considering Definition 3.13 of the CSM and the loss component as well as setting
C
Zt−1 = λt−1 Ft−1 (1 − dt−1 ), we obtain the result
xt−1 = wt−1 + ∇t F U − ∇t ABE + aCSM R
t−1 + xt−1 + `t

by considering four cases:

Case 1: Zt−1 = LLoss
t = 0. In this case, ∆t LCSM = wt−1CSM
− aCSM
t−1 − ∇t T
BE
− ∇t T R
Loss
as Lt−1 = 0. Substituting this result into (7.1) gives
xt−1 = wt−1 + ∇t F U − ∇t ABE + aCSM R
t−1 + xt−1

which yields the result since `t = 0.

Case 2: Zt−1 = 0 and LLoss
t > 0. In this case, ∆t LCSM = −LCSM Loss
t−1 as Lt−1 = 0.
Moreover, `t = −LCSM CSM
t−1 − wt−1 + ∇t T
BE
+ ∇t T R . Substituting this result into (7.1)
gives
xt−1 = wt−1 + ∇t F U − ∇t ABE + xR t−1 + `t
which yields the result since aCSM
t−1 = 0.

Case 3: Zt−1 > 0 and LLoss

t = 0. Then, ∆t LCSM = −aCSM
t−1 − Zt−1 − ∇t T
BE
− ∇t T R
as LCSM
t−1 = 0. Moreover, `t = −Zt−1 . Substituting this result into (7.1) gives

xt−1 = wt−1 + ∇t F U − ∇t ABE + aCSM R

t−1 + xt−1 + `t
CSM
which yields the result since wt−1 = 0.

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293
 THE IFRS17 GUIDE FOR THE PERPLEXED ACTUARY 17

Case 4: Zt−1 , LLoss

t > 0. Then, ∆t LCSM = 0 as well as `t = ∇t T BE + ∇t T R .
Substituting this result into (7.1) gives
xt−1 = wt−1 + ∇t F U − ∇t ABE + xR
t−1 + `t

which yields the result since aCSM CSM

t−1 = wt−1 = 0. 

References
[1] J. Hirz, Advanced Conditional Risk Measurement and Risk Aggregation with Applications to
Credit and Life Insurance, Ph.D. thesis, TU Wien, 2015. 8
[2] M. Kalkbrener, An axiomatic approach to capital allocation, 2002. 8
[3] Long Teng, Matthias Ehrhardt, and Michael Günther, Modelling stochastic correlation, Journal
of Mathematics in Industry 6 (2016), no. 1. 14
[4] B. Widing and J. Jansson, Valuation Practices of IFRS 17, Master’s thesis, KTH Royal
Institute of Technology, 2018, Cooperation with Willis Towers Watson. 2
[5] D. Williams, Probability with Martingales, Cambridge Mathematical Textbooks, Cambridge
University Press, Cambridge, 1991. 2
[6] M. V. Wüthrich and M. Merz, Financial Modeling, Actuarial Valuation and Solvency in
Insurance, Springer-Verlag Berlin Heidelberg, 2013. 3

Electroniccopy
Electronic copy available
available at:
at: https://ssrn.com/abstract=3222293
https://ssrn.com/abstract=3222293