World Bank Document

Produced by the Research Support Team

Abstract

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

Policy Research Working Paper 5377

Why should countries buy expensive catastrophe

insurance? Abstracting from risk aversion or hedging

motives, this paper shows that catastrophe insurance

may have a catalytic role on external finance. Such effect

is particularly strong in those middle-income countries

that face financial constraints when hit by a shock or in

This paper—a product of the Brazil Country Management Unit and the Office of the Chief Economist, Latin American

and the Caribbean Region—is part of a larger effort in the department to understand how developing countries can deal

with natural disasters more effectively. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.

org. The authors may be contacted at tcordella@worldbank.org, and ely@utdt.edu.

its anticipation. Insurance makes defaults less appealing,

relaxes countries’ borrowing constraint, increases their

creditworthiness, and enhances their access to capital

markets. Catastrophe lending facilities providing “cheap”

reconstruction funds in the aftermath of a natural disaster

weaken but do not eliminate the demand for insurance.

Introduction

All countries are exposed and vulnerable to natural disasters.

Exposure is mostly determined by geographic

characteristics, vulnerability by policies, and depends on levels of economic development. Indeed, richer

countries have more resources to …nance risk mitigation activities in preparation for, or reconstruction

activities in the aftermath of, a shock. This may well explain why the cost of the recent terrible earthquake

(both in terms of human lives and economic costs) was much larger in Haiti than in Chile and this despite

the fact that the latter earthquake was stronger.

If countries di¤er in their exposure and vulnerability to natural disasters, should they deal with them

di¤erently? What are the pros and cons of purchasing catastrophe insurance versus doing nothing and/or

relying on ex-post lending facilities such as the ones o¤ered by international …nancial institutions? To answer

these important questions, this paper presents a simple model that sheds some new light on the main trade-

o¤s that countries face in making such di¢ cult choices.

It is well known

that market insurance, if reasonably priced, is the most e¤ective way for risk adverse

agents to cope with large and rare events. This is true for individuals as well as for sovereigns. Indeed,

there are a few recent examples of countries that insured themselves against natural catastrophes:

In May

2006, the Mexican government issued a US 160 million dollar parametric catastrophe bond to …nance rescue

and rebuilding in the case a major earthquake hits some densely populated areas of the country;

in 2007,

a pool of Caribbean countries developed a Caribbean Catastrophe Risk Insurance Facility (CCRICF) that

facilitates their access to the insurance market. However, catastrophe insurance is expensive and even in the

most successful cases, such as the two we just discussed, the cost of coverage turned out to be a multiple of

the fair price (around three times in either case).

Why is insurance so costly? Several reasons are invoked, including supply-side constraints induced by

either agency costs or adverse selection, problems of information opacity of tail events, coordination failures,

and oligopolistic practices. While the securitization

of catastrophic risk through the issuance of catastrophe

bonds may in the future induce greater market discipline, until now it has fallen short of reducing the costs

of insurance to actuarially fair levels.

If insurance is so expensive, why do countries still buy it? One reason could be risk aversion; another

could be the presence of concavities in the production function and/or convexities in the borrowing cost

function that create hedging opportunities as in Froot et al. (1993). While both these assumptions may play

an important role for the demand of insurance, nonetheless we think that they only partially justify why a

By exposure we denote the probability of being hit by a natural disaster and by vulnerabilty the expected loss associated

with any of such disasters.

See Elrich and Becker (1972).

See Hofman and Bruko¤ (2006) for a survey of the insurance opportunities available to developing countries against natural

disasters.

This was the the …rst tranch of a 450 milion US dollars insurage coverge plan. Payment are triggered if a earthquake of

magnitude 7.5 or 8 hits some prede…ned zones of the country. See Nell and Richter (2004) for a discussion of the of parametric
insurance, that is of insurance policies with payments linked to measurable events such as the magnitude of an earthquake, or
the wind-speed of an hurricane.

Note that a loading equal to two to three times the fair price is considerably lower than industry averages that range from

5 to 6 times expected outlays.

See Ibragimov et al. (2008)

For a comprehensive discussion of the market for catastrophe risk, see Froot (2001).

For a discussion of the securitization of catastrophe risk and the development of a catastrophe bond market, see Doherty

(1997).

country may decide to rely on expensive insurance. Indeed, risk aversion does not necessarily transfer from

individuals to countries and the production smoothing argument may not be always a critical one in dealing

with fat tails events such as natural disasters.

For the above reasons, in our model we assume that (i) agents are risk neutral, to abstract from consump-

tion smoothing motives, and (ii) that the premium requested to insure the stock of infrastructure against a

natural disaster is higher than the expected return of rebuilding the same infrastructure, to rule out hedging

motives. In such a setting, demand for insurance arises because of its “catalytic” role on external …nance.

By this we mean that by guaranteeing resources that limit economic contraction in the aftermath of a shock,

insurance makes default relatively less appealing. This relaxes a country’s borrowing constraint, increases

its creditworthiness, and enhances its access to capital markets. For large rare events such as the ones we

consider here, such a benign e¤ect may well outweigh cost considerations.

Of course, not all countries bene…t the same from catastrophe insurance. The ones that bene…t the

most are those medium-income countries that have limited access to the international capital market and

face …nancial constraints either when they are hit by a shock or in its anticipation. Conversely, catastrophe

insurance should not appeal to either poor countries without access to capital markets, that cannot pro…t

from such catalytic e¤ect, or to rich countries that preserve market access even in the aftermath of a large

disaster and thus have no reason to rely on expensive insurance.

Even in those cases in which it is bene…cial for a country, one may nonetheless wonder whether insurance

is an e¤ective choice for low-income catastrophe-prone economies that are likely to receive (belated but

cheap) o¢ cial loans in the aftermath of a catastrophe. To explore whether such donors’Samaritan dilemma

may hinder the demand for insurance, we introduce a multilateral catastrophe lending facility guaranteeing

access to reconstruction funds at the risk-free interest rate in the event of a natural disaster. Unlike private

lenders, multilateral lenders are assumed to enjoy a preferred creditor status that allows to “ignoring”credit

risk.

The introduction of the facility weakens but does not eliminate the demand for insurance because of

the catalytic e¤ect we just mentioned. While insurance entails a positive transfer after the shock (hence, its

catalytic e¤ect), the repayment of the multilateral loan tightens the borrowing constraint, crowds out private

lending, and reduce the country’s investment opportunities in good times. This in turns implies that the

more credit constrained a country is, the greater the drawback of the lending facility relative to catastrophe

insurance, and the larger the demand for the latter.

The paper is organized as follows: Section 2 presents the model and characterizes the benchmark case.

Section 3 introduces and discusses catastrophe insurance. Section 4 does the same for the multilateral

catastrophe lending facility. Finally, Section 5 discusses the …ndings from a welfare perspective and Section

6 concludes.

As documented by Jeanne and Zettelmeyer (2003) for the IMF, multilateral lending to middle income countries is virtually

default risk-free. However, for our purposes it is su¢ cient to assume that the associated default costs are higher than for private
claims.

1 0

To our knowledge, these issues have not been yet examined in the economic literature. By contrast, there is growing

economic literature assessing the economic costs of natural disasters. See, inter alia, Mauro (2006), Ramcharan (2005), Toya
and Skidmore (2006).

The model

Consider an economy endowed with a two-factor Leontief technology:

Q =

minfminf1; Lg; Kg

(1)

to produce a single consumption good. The …rst factor, denoted by L, can be thought of as infrastructure,

which we assume to be at its maximal level L = 1 at the beginning of the production cycle (t = 0). The

second factor, K, represents installed productive units (or capital, for short), which we assume to be zero

at t = 0, and needs to be externally …nanced (see below).

(

> 1) denotes a total factor productivity

parameter.

The timing of the model is as follows:

At time t = 0, the country issues bonds for an amount D

to …nance capital investment K = D

. The

gross borrowing cost i is assumed to be equal to the risk free rate r

(which we normalize to 1 without loss of

generality) plus a risk premium ; itself a function of the probability that the country defaults on the bond

We thus have that i = 1 + .

In the interim period,

t = 1, with a probability

–which we assume to be “small enough” throughout

the paper–the country is hit by a natural disaster that destroys a fraction

> 0 of its infrastructure. Faced

with such a negative shock, the country has the option to issue new bonds for an amount D

to …nance

infrastructure reconstruction, so that L = 1

+ D

At the end of the production cycle, t = 2 output is realized and consumption takes place.

Denoting by the subscript b and g “bad” and “good” states of nature, according to whether the shock

occurs or not, output X in period 2 can be written as:

= x + Q

= x + minf1; D

(2)

= x + Q

= x + minf1

+ D

; D

(3)

where x, x 2 <

, denotes period 2 endowment which we take as a proxy for the country’s income level.

The country’s ability to raise new funds D

; after su¤ering the adverse shock, depends on its access to

capital market, which depends on its creditworthiness and thus on its perceived willingness to repay its

debt obligations. Speci…cally, following the “old” sovereign debt literature à la Cohen and Sachs (1986), we

assume that a default causes the country to lose a share

< 1 of its current output X, a loss that is not

fully appropriated by the lenders. For simplicity, and without great loss of generality, we then assume that

no part of this lost income accrues to the latter.

The country thus faces two distinct borrowing constraints depending on whether default is avoided

altogether, or it is expected only in the event of an adverse shock. In the …rst case, the constraint requires

that default costs, in bad states, exceed the cost of servicing the debt, or:

+ D

(x + minf1

+ D

; D

g) :

(4)

1 1

For the sake of simplicity, and without great loss of generality, we assume that the interim period is close enough to the

initial period so that the borrowing costs are the same in both periods.

Of course, if the country does not default in bad states it would a fortiori not default in good ones. In the

second case, instead, (4) does not hold, lenders anticipate default in bad states and charge a risk-adjusted

interest rate i =

. The borrowing constraint that ensures that default is avoided in good states can then

be written as:

(x + min

f1; D

g) :

(5)

Finally, we assume that

;

(6)

where the …rst inequality implies that investment increases default costs by less than it increases debt so

that a country without endowment has no access to …nance. The second inequality, instead, ensures that

investing in period 0 is always optimal.

In our model, consumers are risk neutral, and policy makers maximize expected income Y ,

(Y ) = (1

+ Y

(7)

where superscript j 2 fd; ndg denotes whether the country defaults if hit by an adverse shock or it does not,
and

;

(8)

;

(9)

Note that in this set-up, income and welfare are mostly determined by the borrowing constraints, and

the latter are, in turn, a function of endowment x. This implies that they are more likely to bind in poor

countries than in richer ones. In fact, given that default costs are proportional to total income, which, in

turn, depends on endowments, the latter plays the role of “implicit” collateral to the bond issuance: richer

countries have more to lose if default is the avenue of choice.

We exploit this dimension in the characterization of the general solution of the “benchmark case” by

distinguishing …ve intervals according to the value of the country’s endowment x. In the main text, we will

provide an intuitive characterization of our main results and we refer the reader to the Appendix for the

complete analytical treatment.

2.1

Benchmark Scenario

In high-income countries,

default costs are large enough to ensure that the borrowing constraint

(4) is never binding. As a result, the country can borrow at the risk-free rate the optimal amount D

= 1

in period 0, and the optimal amount D

in period 1, if it is hit by a shock. Production is always

maximized, and so is expected income.

In less rich countries (x

< x < x

), endowment no longer provides enough “collateral” to ensure that

1 2

The exact values of the thresholds are provided in the Appendix.

borrowing constraints are always slack. As a result, the country cannot borrow D

= 1 in period 0 and

in the event of a shock. In this case, policy makers have a choice between maximizing period 0

investment or “underinvesting” initially, that is choosing a D

< 1 in order to “save” additional access to

…nance should an adverse shock occur in period 1. We can show that, if the shock is rare enough–so that

the underinvesting “self insurance” option is extremely expensive in terms of lost consumption–the country

always chooses to maximize period 0 investment at the expense of period 1 additional access to …nance in

the event of an (unlikely) adverse shock. We thus have that D

= 1, D

< 1

If income further decreases (x < x

), the borrowing constraint now prevents the country from …nancing

the optimal amount of capital in period 0 and avoiding default in bad states. As before, for rare shocks, the

country chooses to maximize borrowing in period 0 at the expenses of borrowing after a shock in period 1,

so that we have D

1, D

= 0. The relevant borrowing constraint (4) now becomes

nd
0

(x + (1

)) < 1;

(10)

where the superscript nd denotes non default in the bad state. Note, however, that the …nancially constrained

country now has the option to increase its indebtedness up to the level that ensures repayment in good (but

not in bad) states. Speci…cally, it can borrow up to (5), which now becomes:

) (x + D

)

(11)

from which

d
0

min

)

x; 1 :

(12)

where the superscript d denotes default in the good state. It can be shown that for su¢ ciently high en-

dowment values (x

x < x

) total income is maximized by the lower (default-free) level of indebtedness,

whereas for lower values of endowment (x

> x

) the country chooses the higher level of indebtedness,

borrowing and investing D

in period 0, and defaulting whenever it is hit by a shock.

Finally, poor countries (x < x

) choose, again, to restrain their borrowing in period 0 so as to avoid

default if hit by a shock in period 1.

The above analysis is summarized in Figure 1.a, where we plot D

and D

as a function of initial income

x, setting the rest of the intervening parameters at reasonable (albeit arbitrary) values.

Intuitively, for

rich countries (x

) creditworthiness is never a problem: endowments provide enough implicit collateral

to ensure access to …nance to exploit investment opportunities in good states, and to fully rebuild the

infrastructure in bad states.

By contrast, all other countries face a trade-o¤ between the amount they can invest in period 0 and what

they can invest in period 1 if they are hit by a shock. For rare events, countries are better o¤ investing more

in period 0, even if this means losing access to …nance in the (unlikely) event of a shock in period 1. In this

context, relatively rich upper middle-income countries (x

) can still (partially) …nance the rebuilding

of infrastructure in period 1. This option is lost to when x < x

Moreover, because for x < x

countries are forced to underinvest in period 0 in order to avoid default

1 3

In particular, we assume that

= 0:02;

= 0:3;

= 1:25;

and

= 0:4

in period 1, they face the choice between borrowing more today and defaulting tomorrow in the event of a

shock, and borrowing less today and avoiding default costs tomorrow; a decision that ultimately depends on

the extent to which investment can be expanded by accepting the higher risk-adjusted interest rate.

In the case of middle-income countries with good access to capital (x

x < x

), the additional

resources do not justify the higher rate, and default is avoided. These resources becomes relatively more

valuable as endowments decline and the …nancing gap widens so that, for lower middle-income countries

> x

), overborrowing is the preferred strategy. However, because of the higher interest rate

charged to the overborrowing country, the …nancial constraint tightens faster in this case (

d
0

nd
0

which, in turn, explains why low-income countries (x < x

) prefer to limit investment to avoid default.

Insurance

The presence of borrowing constraints–that limit initial investment, the post-shock rebuilding e¤ort, or both–

create e¢ ciency losses in all but the richest countries. The natural arrangement to mitigate the problem is

an insurance contract that, in exchange of a premium, pays o¤ the country in bad states. To discuss the

e¤ect of the availability of such an insurance policy, we assume that in period 0 the country has the option

to purchase an insurance contract that pays o¤ an amount Z in the interim period in case the country is

hit by a negative shock. To buy one unit of insurance, the country pays a premium ' =

is the net

present value of a unit expected insurance outlay, and

is a margin that re‡ects, inter alia, intermediation

costs and a risk premium (alternatively, the insurer’s cost of capital, including potential increases if the event

materializes). We assume that the insurance premium is paid up front and …nanced through debt issuance.

In addition, we also assume that

;

(

;

(13)

a non-empty interval for small enough

( < ), to explicitly model the fact that insurance is expensive

but not so expensive that the country would ever buy it. Furthermore, to simplify the analysis, we also

assume that

which guarantees that no default occurs in the insurance case.

The expected income of a country that in period 0 borrows D

+ 'Z to invest K

and to purchase Z

units of insurance is

E(Y ) = x + (1

min fK

; (1

) + Z + D

+ 'Z + D

) :

(14)

With this new instrument available, we revisit the country’s choices as a function of income levels. In the

case of rich countries (x

) no insurance is needed to attain the optimum (the borrowing constraint is

1 4

This stylyied contract applies more directly to the case of a parameterized CAT bond with principal Z and coupon ',

which, in the case of a veri…able natural disaster, virtually eliminates the need for costly state veri…cation. Standard CAT
insurance, by contrast, are typically based on actual losses and cover pre-speci…ed layers, de…ned by a deductible or “retention”
(below which no loss is covered) and a “limit” above which no loss is covered.

1 5

Such assumption does not alter the qualitative results of our analysis (insurance always reduces the likelihood of default)

but limit the number of cases that we have to analyze.

not binding). Moreover, because

> 1, the e¤ective cost of insurance exceeds that of international capital,

and no insurance is purchased. The solution is then identical to benchmark case.

For lower values of x (x

> x

) the borrowing constraint limits period 1 borrowing. As in the

benchmark case, for rare events the country always chooses to maximize period 0 investment (which in this

case attains the optimal D

= 1), so that period 1 is the residual variable. In this regard, insurance plays a

complementary role: by ensuring the availability of resources in the aftermath of a shock, it increases output

in bad states and, through this channel, relaxes the borrowing constraint (4) that becomes:

Z) + D

(x + ((1

) + Z + D

))

(15)

so that

nd
1

(Z; D

)

(x + (1

))

+ (

)

;

(16)

and

(17)

This implies that insurance has a “catalytic” e¤ect on private lending: by purchasing insurance the country

enhances access to the international capital market in bad states. This e¤ect explains why the country

might be willing to purchase insurance even if it is expensive relative to capital markets. More precisely, the

derivative of expected income on insurance is given by:

@E(Y )

(

) + (

> 0

(18)

where the expensive premium (the …rst RHS term) is counterbalanced by the positive catalytic e¤ect (the

second RHS term). On the other hand, given its costly nature, the country would purchase insurance only

as a complement to market funds, so that

nd
1

;

(19)

since increasing coverage beyond Z would simply substitute expensive insurance for less costly debt. In

this way, insurance …lls in for private markets, both crowding in additional private funds and providing the

resources that the market does not o¤er to allow the country to insure against the shock.

For lower income countries (x < x

), insurance will no longer have a catalytic e¤ect, as the country

cannot borrow in period 1: However, the country may still be interested in buying insurance. Why? To relax

its borrowing constraint in period 0, which now becomes

(x + minfD

; (1

) + Zg);

(20)

from which we have that:

nd
0

(x + (1

)) + (

)Z:

(21)

It is easy to check that for small values of

nd
0

> 0 and

@E(Y )

> 0. In other words, since insurance

increases access to capital market in period 0, the country fully insures, that is, purchases insurance for an

amount Z = D

) so as to bring infrastructure to L = K in the event of a shock.

It is important to note that it is never in the best interest of the country to overborrow at the risk of

defaulting in bad states and purchasing insurance. The intuition is the following. Consider a country that

overborrows and defaults in bad states; the anticipation of default eliminates the catalytic e¤ect of insurance

(through its role in the now irrelevant (4)). In turn, without this e¤ect, insurance is simply too expensive

and no insurance would be purchased if default in bad states is inevitable. Note that insurance does not

work as a substitute for lending. Rather, it pays o¤ to insure only if the country reaps the crowding-in

bene…ts, for which insurance funds need to work as collateral to prevent default in bad states. If default in

bad states is anticipated, the collateral value of insurance vanishes.

Finally for poor countries (x

> x), access to insurance no longer plays a role as the catalytic e¤ect

disappears because the lack of creditworthiness limits investment opportunity in period 0.

The previous analysis is summarized in Figure 1.b, where we plot debt and insurance outlays under

the insurance case for the same parameter as in Figure 1.a. In addition, we assume that the overhead

parameter

= 1:3.

As we already pointed out, for high income countries the borrowing constraint does not bind, no insurance

is purchased, and the results are the same as in the benchmark case. Less rich countries (x

> x

instead, start purchasing insurance to increase their ability to rebuild infrastructure in the aftermath of the

shock. Notice that in this case insurance plays two roles. On the one hand, by ensuring the availability of

“pre…nanced” reconstruction funds, it increases output in bad states and, in turn, default costs, crowding

in private lenders. As a result, in this region, insurance enlarges the amount of resources available in the

aftermath of the shock relative to the benchmark, and complements private funds.

Also, the more credit

constrained a country is, the more insurance it buys. Relatively poorer countries (x < x

), instead, are less

a¤ected by the shock (because of their limited capital investment in period 0)), and this negatively a¤ects

their demand for insurance.

A catastrophe lending facility

The previous analysis focused on the “supply side”of the problem showing that the availability of insurances

allows a country to relax the borrowing constraint–albeit at or despite a considerable cost. However, because

large events of a systemic nature such as natural disasters involve massive economic losses and a¤ect a large

number of people, they typically trigger ex-post government intervention which agents correctly anticipate.

This often creates Samaritan dilemma type of problems that lead individuals to underinvest in catastrophe

insurance. Similarly, at the international level, catastrophes in low-income countries elicit an almost immedi-

ate reaction by the international community in the form of (often concessional) loans for social expenditure

and reconstruction. Why would a country bear the exorbitant insurance premiums if it is likely to have

access to o¢ cial resources at a small cost? Is this version of the Samaritan’s dilemma what is behind the

scarcity of catastrophe insurance in middle- and low–income countries? Would it be there any demand for

insurance should the latter be made available to all?

We can easily adapt our model to look into this issue and examine whether insurance is still purchased

by the country in presence of a catastrophe lending facility o¤ering unlimited funds at the risk-free rate in

the event a shock. For expositional purposes, it is easier to tackle this question in two steps, solving …rst for

the lending facility in the absence of insurance, and then introducing the insurance.

Consider now the case in which a multilateral lender o¤ers a catastrophe lending facility, as in the case of

the recent approved World Bank CAT DDO facility,

from which a country can draw (only) in the event of

a shock. It is easy to show that this facility cannot be o¤ered by private markets because loan amounts will

be restricted by the borrowing constraint in exactly the same way D

was in the previous case. However, a

multilateral lender could in principle exploit its preferred creditor status to provide access in period 1 beyond

what the borrowing constraint allows. Indeed, preferred o¢ cial creditors (the government at the national

level, multilaterals and donors at the international level) are the ones that usually come to the rescue after

large natural disasters.

In order to represent the preferred creditor status of the multilateral lender, we assume that defaulting

on the multilateral is prohibitively costly so that multilateral loans are always repaid. Therefore, in this

case, selective default on private creditors could be an equilibrium outcome.

Given the debt D

(with private lenders) and M (with o¢ cial lenders), in period 2 the country faces two

choices: repay or default on bonds.

More formally, in period 2; the sovereign does not default on bonds if, and only if

nd
0

(M ) =

[x + (1

+ D

+ M )]

(22)

whereas the unconstrained M is set to maximize period 2 output, i.e., M = D

Replacing M into (22), we obtain

nd
0

(x + 1

)

(

(23)

Note that the fact that M is chosen ex post (i.e., the country cannot commit not to borrow from the

facility in period 1) simpli…es the problem, which now boils down to the choice of period 0 borrowing, D

Also note that, under the assumption that multilateral and private lending command the same interest rate,

the actual composition of period 1 lending is immaterial for the current analysis. Then, without loss of

generality, we can set D

= 0.

In the non default case, expected income can be expressed as

E(Y ) = x + (

1) D

nd
0

) :

(24)

However, the country can also borrow beyond the limit imposed by (23) and, after a shock, withdraw from

the facility and default on the bond. In this case, expected income is given by

E(Y ) = (1

) x + (

1) D

d
0

) (1

)

(25)

As before, the equilibrium can be characterized by income levels. In the case of rich countries (x > x

)

1 6

The CAT DDO is a new …nancial product o¤ered to middle-income country governments by the International Bank for

Reconstruction and Development (IBRD), part of the World Bank Group.

1 7

We implicitly assume that the multilateral has no way of conditioning its lending on the continued service of the debt with

private lenders, which in our setup simply re‡ects a sequencing issue: the fact that the contingent loan is disbursed in period
1

, before the bond matures. However, we come back to this point in the …nal section.

the borrowing constraint is not binding: the country borrows and invests D

= 1 in period 0, and D

= ,

in period 1 in the event of a shock. However now, less rich countries x 2 [x

; x

], can borrow from the

facility as much as they need because defaulting on it is not an option.

In the case of poorer countries, (x 2 [x

; x

]), period 0 borrowing constraint binds: D

< 1 and, as a

result, M = D

) < . For lower level of endowments (x 2 [x

; x

]) the credit constrained country

chooses to borrow D

= min

) x

1 (1

)

; 1

in period 0 at a risk-adjusted rate i =

, and, if hit by the

shock, borrows M

= D

) from the contingent credit line in period 1, and defaults in period 2

on its obligation to private creditors. The intuition is similar to that in the benchmark case, except that

now the overborrowing country still has access to …nancial resources in period 1. Indeed, overborrowing also

increases output in bad states, since reconstruction funds are not restricted by the borrowing constraint and

increase linearly with period 0 investment. For this reason, default has a smaller impact on income than in

the benchmark. Finally, for the same reason as before, low income countries (x 2 [x

; x

]) choose to avoid

default and borrow from the facility as long as they face capital needs in period 1

A visual comparison of how the multilateral facility (see Figure 1.c) compares with the benchmark case

reveals that the presence of the contingent credit line narrows the interval in which the country chooses to

default. This is so because the contingent credit line increases the value at stake in case of a default. Given

that default costs in this setup are proportional to output, the benign output e¤ect of the contingent line

increases the cost of defaults and reduces their incidence–even though the defaulting country still has access

to the multilateral loan.

In turn, comparing with the insurance case, borrowing in period 0 is never higher under the lending

facility. Again, the intuition is relatively straightforward: whereas the insurance premium entails a transfer

from good to bad states (and, in particular, is arbitrarily small for rare events), the catastrophe loan transfers

the cost of the shock intertemporally within bad states (that is, states marked by the occurrence of the shock),

creating a sharp asymmetry between good and bad states, and tightening the borrowing constraint associated

with the latter. Hence, the lower borrowing amounts (due to the crowding out of period 0 bond borrowing

by period 1 multilateral lending) and the positive probability of default.

Regarding this point, note that for simplicity we assumed that the lending facility extended one-period

loans. While this realistically re‡ect the short-run nature of most emergency and concessional lending,

it bears the question of whether a longer loan can substitute insurance in those cases in which, because of

market imperfections or political economy reasons, supply or demand for insurance is likely to be insu¢ cient.

More speci…cally, can a 1 in 30 years event be covered indistinctly by insurance and by a 30-year contingent

loan?

According to the previous analysis, it cannot. A country that optimally borrows from the facility after

it is hit by a shock inherits the full stock of debt, irrespective of the duration of the loan. In other words,

since default in this case is not the result of a liquidity crisis but rather the consequence of a cost-bene…t

analysis, it is the stock of debt rather than its ‡ow cost that determines the decision.

Consider now the case in which the country has access to both insurance and the lending facility. Would

the country still purchase insurance in this case, or would it rely entirely on catastrophe lending? In other

words, does the facility make the supply of insurance redundant for the country?

To answer the question, …rst note that for relatively rich countries (x

2 [x

; x

]), borrowing from the

facility is clearly superior to insurance, because it allows the country to circumvent the borrowing constraint

at a lower cost. On the other hand, it is easy to verify that, for x < x

, insurance is always demanded.

In particular, for x 2 [x

; x

], see Figure 3a, the country’s problem consists in investing L

= 1, and

, at the lowest cost, which in turn implies minimizing the amount of (costly) insurance compatible

with that objective. The borrowing constraint (4) now becomes:

(x + ((1

) + Z + M ))

(26)

and substituting

M = D

)

(27)

(26) can be rewritten as:

(x + 1

+ Z)

(

(28)

It is easy to verify that within this interval, both insurance and multilateral lending coexist. In addition,

for x

, low period 0 investment levels make the lending facility redundant, and insurance becomes the

only source of funding in bad states. Thus, we are back to the insurance case discussed previously.

In summary, the demand for insurance is crowded out by the presence of the facility only for those

relatively high income countries for which the facility is enough to lift the borrowing constraint. However,

because multilateral lending crowds out access to capital markets in period 0, insurance still plays a helpful

role reducing the burden of period 1 debt, thereby relaxing the borrowing constraint. In other words, while

the Samaritan’s dilemma considerations eliminate the need for insurance as a source of funds in bad states,

it does not eliminate its catalytic role in good states.

Welfare analysis

So far, we have concentrated on the implications of the models in terms of access to …nance in both states.

Naturally, there is more to this exercise than simply comparing access. Indeed, an evaluation of the di¤erent

alternatives would have to ponder their consequences in terms of expected income which, without great loss

of generality, we perform graphically for the set of parameters used in the …gures. Our welfare analysis is

summarized in Figure 2. The top panel (2a) of the …gure plots net income from production in good and

bad states (that is, output net of borrowing costs and endowments, or Y

g;b

x), for each of the three main

scenarios under study: the benchmark, up-front insurance and ex-post catastrophe lending. The second

panel (2b) does the same for expected income. Not surprisingly, both insurance and catastrophe lending are

(weakly) superior to the benchmark: income under each alternative (and in both states) is always greater or

equal than in their absence. But their relative bene…ts di¤er according to the country’s endowment.

If access to …nance is not critical (richer countries), the insurance option yields a lower expected income

than the less expensive multilateral lending facility. However, in the case of low-middle income countries the

multilateral facility may, at the same time, crowd out private lending and be ine¤ective in avoiding costly

default. Since for these countries access to …nance is critical it is not surprising that higher levels of expected

income are associated with the insurance option. What is somehow more surprising is that for a large set

of endowment values a country may enjoys higher income in both states of nature if it relies on insurance

rather than on the multilateral facility.

Such welfare trade-o¤s are clearly illustrated in Figure 3, where we cast a closer look at the situation

in which insurance and the catastrophe lending coexist. As can be seen, the demand for insurance kicks

in at the endowment level for which the borrowing constraint starts limiting investment in period 0. Thus,

by crowding in private lending in period 0, insurance enables a …nancially constrained country to reach the

optimal level of investment, albeit at a premium that detracts from the optimal expected income.

A potentially undesirable characteristic of the lending facility examined above is that it involves a mul-

tilateral institution lending to a country at a time when the country is expected to default on its private

creditors. Unlike implicit arrangements, an explicit facility could still condition access to the facility ex ante,

so as to make sure that the borrower has the incentives to avoid default.

This is not far from standard

multilateral practice: multilateral loans are often granted provided that the recipient country meets certain

debt sustainability criteria.

In this way, the o¢ cial lender ensures that the country does not take the new

money the minute before it defaults on third parties. Intuitively, to the extent that overborrowing excludes

the country from the facility, this new condition should detract from the incentives to default, and reduce

its incidence.

The solution for a contingent catastrophe lending facility (contingent on not defaulting on the private

sector) does not di¤er much from the one presented in the previous section (see Appendix). Interestingly,

a comparison between the contingent and the uncontingent facility reveals the latter to be better, at least

in terms of expected income, see Figure 4. The reasons is that, for those endowment levels for which the

two di¤er, the contingent facility saves the default costs at the expense of leaving the country under…nanced

after an adverse shock. However, because the shock is exogenous, the situation involves no moral hazard

and no value is created by reducing the incidence of default. On the contrary, the punishment (exclusion

from the facility) translates in a lower overall welfare.

Conclusion

In this paper, we have analyzed how countries with di¤erent degree of access to credit markets can cope

with natural disasters. More precisely, we have shown that insurance, even when very expensive, can be

an e¤ective instrument to improve a country’s creditworthiness. We have also shown that the presence of

a credit line that provides automatic access to reconstruction funds ex-post, at a cheaper cost, does not

completely crowd out the demand for insurance.

The previous discussion, however, needs to be quali…ed in three ways. First, the insurance and the credit

line di¤er in one crucial aspect: the loan has to be paid after a bad shock while the cost of insurance is

transferred to good states. In other words, income (and consumption) volatility is bound to be lower with

insurance than in any other scenario. Trivially, if income smoothing–from which we deliberately abstracted

so far–were a policy objective, insurance would become relatively more appealing, a result that would only

add to the case for insurance that our …ndings support.

Second, we assumed that the ex-post credit line is not concessional and that the lender recoups the cost.

In reality, often disaster help is highly concessional. Our main results would hold true for a limited level of

1 8

To enhance incentives without distorting its automatic nature, the facility could involve temporary subscriptions on a rolling

basis, to ensure that the country is not cut o¤ overnight but still faces frequent exams.

1 9

However, in the aftermath of a natural catastrophe it is not unlikely that other creditors (especially bilateral) agree to

write-o¤ part of their credit to allow multilateral lending.

References

[1] Becker, T.and P. Mauro (2006), “Output Drops and the Shocks That Matter," IMF Working Paper

06/172

[2] Cohen, D., and J. Sachs (1986), "Growth, External Debt, and Risk of Debt Repudiation," European

Economic Review, Vol. 36, pp.687-93

[3] Doherty, N. (1997), “Innovations in Managing Catastrophe Risk,” The Journal of Risk and Insurance,

Vol. 64: pp. 713-718.

[4] Ehrlich, I. and G.Becker (1972), “Market Insurance, Self-Insurance, and Self-Protection,” The Journal

of Political Economy, Vol. 80, pp. 623-648.

[5] Frooth, K. (2001), “The Market for Catastrpohe Risk: a Clinical Examination,” Journal of Financial

Economcs, Vol. 60, pp. 529-71.

[6] Hofman, D. and P. Bruko¤ (2006), “Insuring Public Finances Against Natural Disasters–A Survey of

Options and Recent Initiatives,” IMF Working Paper 06/199

[7] Ibragimov, I., Ja¤ee, D., and J. Walden (2009), “Nondiversi…cation Traps in Catastrophe Insurance

Markets, The Review of Financial Studies, Vol.22, pp. 959-93.

[8] Ja¤ee, D and T. Russell (1997), “Catastrophe Insurance, Capital Markets, and Uninsurable Risks," The

Journal of Risk and Insurance, Vol. 64, pp. 205-230

[9] Jeanne, O. and J. Zettelmeyer (2001), “International Bailouts, Moral Hazard, and Conditionality,”

CESifo Working Paper Series No. 563.

[10] Nell, M., and A. Richter (2004), “Improving Risk Allocation Through Indexed Cat Bonds,”The Geneva

Papers on Risk and Insurance, Vol. 29, pp. 183-201.

[11] Ramcharan, R. (2007), “Does the exchange rate regime matter for real shocks? Evidence from wind-

storms and earthquakes," Journal of International Economics, Vol. 73, pp. 31-47.

[12] Toya, H. and M. Skidmore (2007), “Economic Development and the Impact of Natural Disasters,”

Economics Letters, Vol. 94, pp. 20-25.

[13] Yang,

D. (2006),

“Coping with Disasters: The Impact of Hurricanes on International Fi-

nancial Flows,

1970-2002,” Unpublished Manuscript,

available on the web at http://www-

personal.umich.edu/%7Edeanyang/papers/papers.html

APPENDIX

7.1

The benchmark case

7.1.1

Case 1B:

Setting D

= 1, and D

= , it is immediate to verify that (4) becomes

1 +

(x + );

and is satis…ed if x

In turn, K = D

= 1, L = D

yield

(Y ) = x +

(1 +

) :

7.1.2

Case 2B:

> x

)

If x < x

, the country cannot borrow D

= 1 in period 0 and still have access to D

in bad states.

Therefore, it faces a trade-o¤ between maximizing period 0 investment, or “underinvesting”initially in order

to “save” additional access in bad times.

Note that in equilibrium L

K =) 1

+ D

so that the relevant borrowing constraint (4) can

be written as:

+ D

(x + ((1

) + D

)) ;

so that:

nd
1

x +

)

In turn, we can express the trade-o¤ between increasing consumption in good and bad states in terms of

the country’s problem at time 0:

max

[Y (D

; D

nd
1

)]

x + ((1

+ (1

))

+ (

1)D

nd
1

subject to 0

nd
1

)

It is then easy to verify that, for

low enough,

@E(Y )

= (1

)

(

> 0;

(29)

which indicates that the country maximizes period 1 investment (D

= 1) at the expense of lower investment

in the event of an adverse shock.

Finally, we need to verify that the country can invest a positive amount in the second period while

avoiding default. This condition can be written as:

nd
1

))

0 () x

B
2

)

(30)

In sum, for x 2 [x

; x

], we have that K

= D

= 1, and, L

= D

x (1

))

< , and

(Y ) = x + (1

) + D

(

7.1.3

Case 3B:

> x

(

1)(1

)

It follows from (30) that, in this interval, D

= 0. Moreover, the country cannot borrow the optimal amount

of capital in period 0 without risking default if hit by a shock.

From (10) and (12),

nd
0

= (x + (1

)) < 1:

(31)

and

d
0

min

)

x; 1 ;

In turn,

d
0

= 1 () x > e

)

(32)

and

()

1 +

;

(33)

so that for a su¢ ciently small

, there exists a non-empty interval [

x; x

] such that D

= 1. We also have

that

nd
0

> (1

) () x >

)(1

)

and

x ()

< 1

)

so that in the interval [

x; x

], minfD

; D

g > 1

so that the countries have an interest in borrowing if

they are hit by the shock.

Finally, we have that:

Y ((D

nd
0

))

)(x + D

nd
0

) +

[x + (1

)]

nd
0

;

(34)

(Y (D

d
0

))

)(x + D

d
0

) +

) [x + (1

)]

d
0

;

(35)

substituting the values for D

and D

from expressions (31) and (32) in these expression we have that:

d
0

; D

nd
0

Y D

d
0

Y (D

nd
0

)

1 + (1

)(

1)(x

);

which is linear in x.

Trivially, for x = x

, we have that D

= D

= 1, and

; D

> 0. On the other hand we

have that

lim

nd
0

; D

d
0

lim

1; D

nd
0

(

)

(

1));

expression that is negative if

is small enough. From this, it follows that within the interval there is

a unique value of x

2 [e

x; x

] such that E(Y (D

))

E(Y (D

)) > 0 , x > x

. It is then easy to

verify that x

(

1)(1

)

) , so that in the interval [x

; x

] the country chooses to borrow

= D

(x + (1

) ) and default is avoided. Income is then given by

(Y ) = (1

) x + D

nd
0

+ (x + (1

nd
0

(36)

7.1.4

Case 4B:

> x

(

1)(1 (1

)

)(

)

From the previous proof, it follows that for

, E(Y (D

))

E(Y (D

)) > 0, and K

= D

= 1.

Consider now the interval [

x; x

] for which:

d
0

)

x < 1;

First of all let’s verify that in this interval minfD

; D

g > 1

. Indeed, we have:

d
a

fx : D

d
0

= (1

)g = (1

)

; and D

d
0

) if x > x

d
a

;

nd
a

fx : D

nd
0

= (1

)g =

)(1

)

; and D

nd
0

) if x > x

nd
a

;

from which we have that

d
a

nd
a

)

> 0

and, in turn, that in the interval [x

; x

], x

> x

=) minfD

; D

g > 1

: Indeed,

B
4

nd
a

)(1

)

) ((1

)

> 0 ()

< 1

which is always veri…ed for

small enough. We can now substitute the values value into D

, and D

(34) and (35) to get

d
0

; D

nd
0

)

(x + (1

)) (

1)) +

((1

)

1) + x (1

))

)

() x

B
4

(

1)(1

)(1

)

)(

In sum, for x 2 [x

; x

];

; D

> 0, and the country chooses to borrow K

= D

= minf

) x

1 (1

)

; 1g

in period 0, and defaults if hit by a shock in period 1. In this case, income is given by

(Y ) = (1

)(x + D

d
0

) +

) (x + (1

))

d
0

7.1.5

Case 5B:

> x

Finally, it is easy to check that for x < x

; K

= D

= (x + (1

)) and the country does not default.

In this case, expected income is given by, in case 3B, by

(Y ) = (1

) x + D

nd
0

+ (x + minf(1

); D

nd
0

7.2

Insurance

7.2.1

Case 1I

(

x > x

)

In the case of rich countries (x

)no Insurance is needed to attain the optimum (the borrowing constraint

is not binding). Moreover, because

> 1, the e¤ective cost of insurance exceeds that of international capital,

and no insurance is purchased.

7.2.2

Case 2I

(

> x

)

As in case 2B, the borrowing constraint determines period 1 borrowing. However, unlike in the benchmark,

insurance plays a complementary role by increasing the collateral and relaxing the constraint. Speci…cally,

we can write the country’s problem at time 0 in terms of D

, D

and Z as:

max

;Z;D

[Y (D

; Z; D

)] = x + ((1

)

1) D

) + (

1)D

(

) Z

(37)

subject to the borrowing constraint

Z) + D

[x + (1

) + Z + D

)]

(38)

from which

nd
1

(Z; D

)

(x + (1

))

+ (

) Z

)

;

(39)

with

nd
1

Substituting (39) into (37), we have that, for any given Z, and for small

(

1) (1

)

) + (

;

(40)

@E(Y )

= (1

)

(

)

> 0:

which implies that the country maximizes period 0 investment and tells us, in particular, that D

> 0 =)

= 1.

Consider the case in which endowment x is large enough to allow the country to borrow D

= 1.

Di¤erentiating (37) with respect to Z we obtain:

@E(Y )

(

) + (

> 0

where the positive sign comes from (13), so that the country purchases insurance subject to

nd
1

(41)

Finally, substituting (41) into (39) it is easy to verify that:

nd
1

; 1

(x + )

(1 +

)

() x

IN
2

and that

1 +

(x + )

)

0 () x

1 +

= x

B
1

so that in the interval x

; x

, D

= 1, D

> 0 and Z

> 0.

7.2.3

Case 3I

(

> x

)(1

)

If x < x

, the country fully exhausts its access to capital in period 0, so the “catalytic”e¤ect of insurance is

re‡ected directly in the borrowing constraint determining access in period 0. Speci…cally, insurance increases

income in period 1 (the value at stake in bad states), at a cost proportional to the likelihood of the insured

event, relaxing the borrowing constraint that now becomes

(x + minfD

; (1

) + Zg);

nd
0

[x + (1

)] + (

) Z:

(42)

Substituting D

in the objective function

E Y D

nd
0

; Z

= x + ((1

)

1) D

nd
0

)

(

) Z

we have:

@E(Y )

0 ()

(43)

which is trivially veri…ed for a small enough

This in turn implies that the country always purchase insurance subject to

= D

);

where Z

denotes the point at which insurance funds allow the country to rebuild infrastructure to the level

L = K.

Finally, substituting Z

in (42) we have that

0 () D

nd
0

x + (1

)

) () x

IN
3

)(1

)

In this interval the country has the option to borrow D

, such that 1

> D

, at a risk-adjusted

rate, and default in bad states. If so, the borrowing constraint (5) becomes:

(x + D

)

(44)

from which

d
0

min

) x

)

; 1 ;

(45)

and expected income modi…es to

Y (D

d
0

; Z) = (1

)x + (1

)

d
0

[ (1

)(1

)

[

)] Z)] :

The …rst thing to stress is that no insurance is purchased if default is anticipated. To see that, note that,

for D

= 1,

@E((Y (1; Z))

[

)] > 0 ()

> (1

);

whereas for D

< 1,

@E((Y (D

; Z))

[ (1

)

]

[ (1

)

< 0 ()

)

(46)

From (13) and (6), we know that

)

, so both conditions hold.

Then, substituting Z = 0 into (45),

d
0

= 1 () x

)

Also, for the default option to be the equilibrium we need that

d
0

= min

) x

)

; 1

> D

nd
0

x + (1

)

;

which in turn implies that

)

= 0:

Therefore, to show that the country never chooses the default option, it su¢ ces to show that it is so for

x =

x, the point at which the additional borrowing in period 0 relative to the non-default case is maximized.

Finally,

d
0

(x =

x) = 1 > D

nd
0

x + (1

)

which would imply

)

;

which is never veri…ed for small

, since

7.2.4

Case 4I

(

> x

)

If x < x

, the country does not purchase insurance and we are back in the benchmark case. To ensure that

no default occurs, it is su¢ cient to show that

IN
3

B
4

)(1

)

(

) +

)

> 0

which is always the case for su¢ ciently small values of

7.3

Catastrophe lending facility

As before, for high initial incomes (x

(

)

), the borrowing constraint (22) does not bind and

the country invests the optimum in both states. In turn, for x < x

, from (23)

nd
0

(x + 1

)

(

and

M = D

nd
0

) =

) (1

)

(

> 0 () x > x

M
4

) (1

)

= x

IN
3

;

(47)

Alternatively, a …nancially constrained country may choose to increase borrowing from private lenders

(at a risk-adjusted rate i =

) at the expense of defaulting on bonds if hit by a shock. However, unlike in

the benchmark, now the country would still have access to multilateral lending in period 2. In this scenario,

period 0 borrowing D

should be such that the country repays in good states:

d
0

) (x + D

d
0

)

from which

d
0

) x

)

1 () x

M
2

)

;

with

for small

, and

d
0

) =

) x

)

() x

M
3

)

> x

M
4

From (24) and (??), we know that the condition for a default on bonds is linear in x:

d
0

; D

nd
0

; M

E(Y (D

d
0

; M

)

E(Y (D

; M

) > 0

) x + D

d
0

) x + (1

+ M

)

d
0

(48)

) x + D

nd
0

x + (1

+ M

)

nd
0

and can be written as

d
0

; D

nd
0

; M

= D

d
0

nd
0

)

[x + (1

)]

(

1) (1

) M

: (49)

As before, we can distinguish three intervals:

; x

where D

= 1 > D

[x+1

]

(

;

where D

) x

1 (1

)

; D

[x+1

]

(

< 1, M

> 0; M

> 0,

;

where D

) x

1 (1

)

< 1

< D

[x+1

]

(

, and M

> M

= 0, and no default

occurs

Given the linearity of (49), to characterize the equilibrium it su¢ ces to check the thresholds for the …rst

two intervals.

Trivially, for x = x

, (49) does not hold and no default occurs, since D

= D

= 1. In turn, for

x =

, D

= 1 > D

implies that (49) always holds for small enough

. Finally, for x =

d
0

M
3

= 1

)

+ (1

) (1

) = D

nd
0

M
3

(50)

and, again, the country does not default.

It follows that there is an interval x

; x

, such that x

> x

, within

which the country borrows D

= min

) x

1 (1

)

; 1

in period 0, and, if hit by the shock, borrows M

) from the contingent loan credit and defaults.

The thresholds for the default interval are obtained directly from (49) as:

M
2

x :

1; D

nd
0

; ; M

= 0 =

(

) (

(

)

and

M
3

x :

d
0

; D

nd
0

; M

= 0 =

) (1

) (

1) 1

(1 +

(

1)) +

(

1) + )

7.4

Catastrophe lending and insurance

For x 2 [x

; x

], the country minimizes the amount of (costly) insurance such that it still attains L

= 1,

and L

= . The borrowing constraint (4) then becomes:

(x + ((1

) + Z + M ))

(51)

which, substituting,

M = D

)

yields

(x + 1

+ Z)

(

M IN
0

(52)

Finally, we have that

M IN
0

1 () Z

M IN

)

;

and, in addition, for M to be non negative, we need

The two conditions are simultaneously veri…ed for

1 +

= x

IN
2

Consider now the case x

. We can now substitute D

from (51) in the expression for expected

income, so that

E(Y ) = (1

) x + D

bl
0

+ (x + (1

+ M + Z)

M )

bl
0

and, di¤erentiating with respect to Z, we get

@E(Y )

( +

) )

(1 +

)

(

;

which is always positive for small enough

. Thus for x

we are back to the insurance case.

7.5

A contingent catastrophe lending facility

The model can be readily modi…ed to represent this case: we simply need to note that no multilateral

assistance is forthcoming in the event of default (M

= 0), which tilts the balance against the default

decision: default, while still possible, is associated with narrower interval.

Under this new assumption, M

= 0, and the borrowing constraint (22) becomes

d
0

; D

nd
0

; 0; M

= D

d
0

nd
0

)

[x + (1

)] + (

1) M

It is immediate to verify that, for x =

, D

= 1,

nd
0

))

(

< 1

and

1; D

nd
0

; M

= 1

nd
0

)

[x + (1

)] + (

1) M

> 0

for small

. On the other hand, for x =

, M

= 0 and we are back in the previous case, where from (50)

we know that the country chooses to borrow less and avoid default.

Thus, following the steps of the previous proof, it can be shown that there is an interval

; x

such

that x

> x

, within which the country borrows D

= min

) x

1 (1

)

; 1

period 0, and, if hit by the shock, defaults.

The thresholds of this interval are de…ned by the zeros of

1; D

nd
0

; M

(x + 1

)

(

1 +

)

(

)

f [x + (1

)]

(

1) (1

and

d
0

; D

nd
0

; M

) x

)

(x + 1

)

(

1 +

)

(

)

f [x + (1

)]

(

1) (1

)g ;

from which

(

1) (1 +

)

(1 + (

1) (

( (1

) + )

))

(

1) (1

)

and

(

1) (1

) (1

)

) ( +

)

( +

(

) (1 + ) +

) (

1)) +

)