Evitoral Policy: ‘The Journal of Ivernaiona! Economic is designed to sare athe prncpa cute or analytes! work inte BE Stintometonal eae ands Ualanca-of-payments snl. and aio or insttutonal, empl, 2d ezcnorete WH Fgh aueliy and goncrl erations Iteart. "wil Seek Co achieve bubnce wong these dfterent sae TPeatonel economia ena an Gee thus aing she widest pofesionslreaderiun mn to eo of Editorial Board Exes Richa A. BRECHER. Dapatment of Econories, Cexleton Unvasny, Orona, Ontario K1S 580, Canada LMgash N BlaGwAr), Depsriont of Enomies,Colimbia Uversiy. Now York NY 10037, US 12fn &CHIPMN, Economics Deparment Unversity of Mnneote, Maneapok, MN GBASS, USA TE BRANSON, Woodrow Witon Schad, Princeton Univesity, Princeton, 08540, USA ‘Rohe cl FEEASPRAY Geounmant of Eoonarca, Unversty of Cains, Da Ga S50%, USA EersrcHELPNran, Beparenent of Ccanoriee ta fow Univesty Tel uw 68878, soe! : Sferer Mooarck Se kalagg Gradusts Scho! of Managemen, Norhwestsn Unversity, Eveston, #60201, Book Bevan Eatar Beet ECHENEREEN, oosatment ot Eoonemies, Unwowty of California, Bukeey, CB $4720, USA Assciate Eft Gatisite EAI, oopanment of Economies, Unieesty of Foneyivans,Fniadeipns, Pa 1908, USA Runaté COOPER, Deparment of Economics, Hevea Unimiy, Cambridge, MA G2¥38, USA on DEARDORFE Desronent of Eonomies, Unwaroey of Michigan, Ann Arbor, MIABYO8, USA Auer DOPRBUSGH, Doponen of Eeanomen E52 257, IT. Combndge, MAOZI3E, USA. “Jonctien EATON, Deparmemt of Eeonamics. Univers of Vitis, Chrcuosvile, VA 2901, USA ‘Woted ETHIER, Dopanman of Econom, Uv of Pemayiania,Phisglphia, PA 1SYO4, USA Foner > FLOOD. Deparment sf Econorsee Norsemen Unversity Evanston, 6620), USA Totus HATTA. Desermant of Pica Eznomy, Johns Hopkins Univesity, Batimor, MD 21218, USA Miuray € KEMP, Schoo’ of Bconomes, Unvaraty sf Nev South Wales, Ressington NSW 2033, Aveta Miat'e LexKER Soosrmont st Economie Unity of Calorie, Los Anges, CA 30028, USA [ater LINDBECK: iste for Interne! Eoonomie Studie, Fach 10405 Stockh 60, Sawdon Jenes Rt MARKUSEN, Deparment of feononiceUrivrty of stn, OntaraLandon, Orato NBA C2 a Miligang MAYER, Doparment ef Eeosomice Univer of Cncinnat, Gres tnai 45221, U Musee bestret, Deparment of fonomice, Unverty of Pearayaria, Piodepis, PA 19104, USA Douglas ©. FURVIS, Depart of Economics, Quee’s Unweraty Kingston, Ontana K7L 2N6, Conads Paula SAMUELSON, Duparmert of Economia, £52-383 M7, Canbndge, MAGZT39, USA [ead SMITH. Senool of Socal Scenes, Unversity of Susex Brighton BN1 GON, UK Me'Zo. SVENSSON, Insite for Iematonal Economie Stade, Univerciy of Stockhoim, S-105 81 Sc Sweden ‘varus UEKAWA, Deparment of Economies Kobe Univesity of Commace, Tami, Kobe 055, Japon ‘Submission of papers: Papen on pobtton shouldbe sont in tipete to Profesor R. A. Biechor_ Deparment of Economie, a GRRE Bea, Once RIS S60. Canags. Tere wo Submission fee uf US #30 fr ll wrod an ‘orto for aubleaton, Cheques #iuid be rade payable to he Jour! of iremaions! Economics. ‘Subscriptions: 1988 (vale, 24 and 25)--800 pages in four quarterly issues ) Joo ictna of ierntional Economies {SSN D0E7-1656) Is Dublshed in woos of to ies, Thea Ite fr 1988 (Volumes Ze and 28) fs Of 810.00+DIL 5200 ppan =DN. 56209, Our ppm (nose a Ravana) charge includes suraco deiner of aliens, excaptto subscib in the folowing count Suey (SALT™ Suc Ai ied) urd grt, Aura, Brn Cnet Cn or Fan lyon” Mayo, Monieo, Now Zealina, Pakean,Seegspote. South Aiea, South Kore, Towa, Tend US aaeeea are mons mel charges are avalale upon request. Reduced Was ot pve auberos ae ol {he Pabienar Cains for mosina Reve wil be onored trae of cnergef made vthin vee months of pubes ‘Sienand Ihave to Couns Senor Publaher BY. (North Halen), Journal Diiwon, P.O. Box 2 SXinstergom, Te Netbrons, (© 1867 ELSEVIER SCIENCE PUBLISHERS 8 V. (NORTH-HOLLAND) B57" AlowMotonsh, Pox HET 100082 Anceraem: The Necwnas Sutraton of ens ubesion arnt we eto iravocale we exc abrzaon of pub ci cece Econo ofaeucian payne te pra as reo mare 17 png 2 ot he OWCh CR SMI aerl Ohewe a he 301800 (8 SN) ice eo te Duh Copyra Ai of 182) wer ‘Serer fay or 108 of he US. Copyott aw, The aporamin te shvle te foward Erente Csrnce Coste. 27 Congo Sst Salem. WA CIBTO USA Ho ace eae ‘oes capt pena mate ned ety feo Att aise stens sn nobis ach unm staeibng tok oat we geri (0 oe "ior ee Te cone dos nt extend wo ata na coe, Journal of International Economies 3 (1987) 201-220. North-Holland TRADE AND INSURANCE WITH MORAL HAZARD Avinash DIXIT* Princeton University, Princeton, NJ 08544, USA Received November 1986, revised version recived March 1987 Previous literature on the use of trade policy for insurance bas relied implicitly on moral hazaré | and adverse selection as the reasons for the failure of private markets for rsk-bearing. This paper begins with the observation that such forces should be incorporated explicit into the | models to ensure a far compariton between markets and government. For the cas of moral Iaaard, it is found that market equilbviam given exclusive insurance is an informationaly constrained Pareto optimum. In one case of non-exclusive insurance, a sjttematic role for ‘overnment insurance is ound, but the scope of trade policy i ambiguous, 1, Introduction It is often argued that in the absence of complete markets for risk-sharing, ‘trade policy can have a second-best role in providing insurance, The most prominent exponents of this view arc Newbery and Stiglitz (1984) and Eaton and Grossman (1985). The former construct an example where autarky is ‘Pareto superior to free trade. The latter examine a more general model using | numerical simulation methods, and find that the optimal policy usually involves some protection, ie. has an anti-trade bias, Both pairs of authors ;ppea] to moral hazard and adverse selection to motivate the absence of ‘Insurance markets. However, they do not include these forces explicitly in This procedure constitutes an unfair comparison between market equilibria | end policy optima. First, the government's conduct of trade policy is = assumed to be free from problems that would plague the market. Second, “neither moral hazard nor adverse selection necessarily implies a total collapse | of insurance markets or other risk-sharing arrangements. _ To make a fair comparison, one should construct a trade model with “moral hazard or adverse selection, allow markets as much play as they can “heigl advice, and the National Science Foundation for esearch support under grant No. SES- {25H Taio hank Noman land and Kale Kins for wel Gomme onan erie eS EEE ELE LEE EEL ELSE EEE2m. A. init, Trade and insurance with moral hazard A. Dist, Trade and insurance with moral hazard 203 have under the circumstances, and then ask if and how policy can impro upon the outcome. The purpose of this paper is to initiate this resear d= F(M)—MF(M), Q) separate forthcoming paper. Some genetal issues of economic policy under moral hazard and ad selection have been examined in this rigorous way, for example by Green and Stiglitz (1986) and Arnott and Stiglitz (1986). They have identified som market failures, and examined how policy can act upon them, My aim ‘more specific; I shall construct a model that is closely related to those fami in trade theory, and see which of the market failures have relevance for it My focus i on the role of trade and tax-transter policies when there it uncertainty with moral hazard. To highlight the issue, I use a model where such policies would have no role in the absence of such problems: a sm cconomy without distributive conflicts. To simplify the analysis, T keep th model small: two sectors, one being safe and the other risky, and one mobile factor of production, labor. The set-up is quite similar to Grossman's (1984), tirade. model of occupational choice, with some inessential details omitted, f assumed to accrue to the government, and then enter consumers’ incomes ia the tax-transfer policies. It would make no difference to give each con- (1984). But that would give equivalent results and merely complicate the “algebra. 3m lump-sum tax t on each worker. Note that t must be determined as | part of the equilibrium, ( and insurance, Risk may be purely individual, or it may have an aggrega (systematic) component, As for insurance, if the total coverage purchased an individual can be observed, then moral hazard is better controlled’ making each insurance contract contingent upon the total coverage for th individual, The effect is just the same as requiring each person to buy all h insurance from the same (competitive) firm. This is called the exclusivity requirement; see Shavell (19/9, in. 8) and Arnott and Stiglitz (1986, pp. 3-4) _ for a full discussion. Thus, we have another choice: to assume exclusivity, of not to, The alternative assumptions about the nature of risk and the kind of insurance give rise to four separate cases. In the next three sections T shall isk, exclusive insurance There are (NM) workers in the risky sector. It is perhaps simplest to | think of them as independent household farms or firms. Each of them jo0ses his level of effort e, measured by utility units, before the uncertainty is resolved. The probability distribution of outcomes depends on e. The EVULP)—e, | where E is the expectation operator. For algebraic simplicity, I shall assume that there are just two possible mutcomes: low output x, and high output x. The probability of the high outcome is xe), an increasing and concave function of e. The probabilities | are independent across individuals. With a large number of workers, the per | capita output is riskless. The important assumption is that the range of possible outcomes, here the | set (x..xu}, does not vary with e. If they did, there would be events whose Occurrence would reveal the chosen e exactly. Then contracts that mete out “sulficently severe punishments in such events could control moral hazard “fully or almost fully, ie. achieve or approximate the full-information first _ best. Therefore the assumption of invariance is the natural one to make in a | model that sets out to examine the problems caused by moral hazard, Otherwise, nothing hinges on the special structure. One could allow a “continuum of possible outcomes x and a probability density function f(x,) them here. The economy has N identical workers. Each has an indireet utility function V(F,p) for goods’ consumption with the following properties is disposable income in terms of the safe good (¥), p is the relative price the good produced in the risky sector (2), and V is increasing and concaye in y Each individual must choose the sector in which to work before any uncertainty is resolved. Let M denote the number who choose the safe sector, ‘There is an increasing concave production function Y determined competitively: w=F(M).204 A. Dist, Trade and insurance with moral hazard at the cost only of analytical complexity; see the discussions in Hart Holmstrom (1986). Similiarly, one could allow hired labor or material inpi making the entrepreneur the recipient of profits, as in Grossman (1 Since the results are unaffected, the simplest structure is the best for purpose. a The government makes net transfers of g1,2n to each X-sector worker wl realizes low and high output respectively. This specification can cover mat policies. If g, >0> gy, we have some publicly provided insurance. If ¢.= 24 ‘we have a reward for undertaking the risky activity. If gy >0=gy. we hay ‘an income-support scheme. a ‘There is private insurance, defined by cach worker's net receipts 2, and: in the respective states. Recall that total coverage can be observed at controlled under the exclusivity assumption of this section. Therefore 1 individual cannot freely choose the scale of coverage at a given ratio of 2 The income of each X-sector worker in the two states is hap te tant [y= Pint Sat ant ‘The expected utility is y= (1-10) 9) + m6) Ul Pe 4 Given the government policies and the insurance contract, the work chooses to maximize ux. Assuming an interior solution (¢>0) the order congition is wleV Up) Vl, pi) The private insurer's zero profit condition is . (=nfe})z, +nfe)zy=0. i Although each person must sign up with just one insurer (exclusivity), t is competition among alternative potential insurers for clients. Therefore insurer will choose (2,24) 10 maximize profit subject to providing competitive level of expected utility ug, and in equilibrium the profit equal zero. It is equivalent and simpler to regard the choice of (, 29) maximizing u, subject to the zero profit constraint (7). The insurers are a aware of the moral hazard, ic. the dependence of ¢ on (2,29) via Therefore we can formally regard them as choosing e as well, subject 10 treated as a constraint on the optimization problem. The Lagrangian for th A, Dist, Trade and insurance with moral hazard 20s He) (bys P) +H) M yp) =e —Agl(l —nte))zx + meen] + 02 (VV ie P)— VL Pd 8) i first-order conditions are (1 —m(e) VAAL P) — Ax] — ae (e) Vill, P)=0, (9) - me) (Vids P)— Ax) + per (e)Vi( a P) = 9, (10) : Agate), —zu) + 1x"(e)[V (Un, P)~ VT BY} = 9. (uy ‘The multiplier for the zero profit constraint is called 2y because it is the “marginal utility of a unit of sure income 10 an X-sector worker | For readers unfamiliar with moral hazard problems, here is a brief “characterization of the solution. First, we must have 1>0. To prove this, “suppose 10. Then (9) and (10) give Vali P)= Axil — ree —n(@))] Sd E VAliy P)=AxiEt + na (e)inteN) & ho Then Vile P)ZVAlisP) 80 Jyh, and Vly,p)SV(l,,p)- Then (6) cannot hold, With >0, we have Vi{l,,p)>4x>Villysp). Then I, 0- “Combining this with (7), we find 2,>0>2_. Thus, some insurance is offered despite moral hazard, For a detailed discussion of these issues, see Shavell (1979). Now we can complete the description of the equilibrium. In the safe sector _¢ach worker has disposable income ay (13)2065 A. Dixit, Trade and insurance with moral hazerd A Dist, Trade and insurance with moral hazard 207 Equilibrium of occupation choice gives bears his (/N}th share of that, Bearing in mind all these elfects, we have @ y= Aity.= gl dg—4g-(N— MY/N-+ ANN] 6) ‘i and It remains to ensure the consistency of the tax policy. Let the consumpt ‘of the X-good by each Y-sector worker be cy, and that by each X-s uy = AyL4w—Ag-(N MN + ATTN], a7) worker, with respectively low and high incomes, be cand cy, In fact Roy's Identity, _ where 2y is the sure marginal utility of income in the Y sector. In the new ty. Let K=Ax/2y, and note that AIT=—M Aw _ by Hotelling’s Lemma. Then (16) and (17) give = VillpPVloP) (= YH. Then total lump-sum tax receipts equal the requirements of the grant of K[ 4g: M/N — Aw: M[N] = dw-(1—M/N)~4g-(1—-M/N). insurance policies in the risky sector, minus the safe sector's profits, minis. Hheraiofex revenuesor a _ This simplifies to Aw=dg, and then (16) and (17) become Au, = duuy=0. Thus, the initial passive policy satisfies the first-order conditions for social ‘optimality. Trade policy is similarly powerless. ‘As usval in optimal taxation theory. e.g. Diamond and Mirrlees (1971), second-order conditions are difficult to verify. However, we can be confident that we have found a maximum by observing that some special cases of the ‘model are familiar ones where the market equilibrium is known to be Pareto optimal. Most simply, if moral hazard is vanishingly small, so n(e) is lat at mt its equilibrium value, then we have a standard Arrow-Debreu model. More equations (thirteen) in (1), (3}-(7) and (9}{15). For my purpose, equilibrium “7 AG" importantly, if the safe sector is vanishingly small, we have Shavell's (1979) is determinate In Shavell’s (1979) model of moral hazard, the competitive equilib with exclusive insurance is constrained Parcto optimal. That is, a s planner facing the same information constraint ~ unobservable effort levels M [FM ~ MF (My) +(N MYLO —2(@))g, + eg 0] saw above. [The intuitive argument above showed how the policy favoring the risky | sector was defeated by the induced migration. This bears out the general a belief that if the government supports risky activities, too much risk will be risky activity is the whole cconomy, and therefore a competitive insurer and Baken, The principle should have wider validity than the trade context problem, Here we have another sector, and labor and funds can be movee between them. The question is whether trade restrictions or public insurance policies can do so in a beneficial way. The answer is no; the competit ‘equilibrium with exclusive insurance remains constrained Pareto optimal EWiéa én, individual total insurance coverage canmot be observed, ex- In the text I shall develop this argument in an intuitive way; appentlix A clusivity cannot be enforced. People are tempted to expand coverage, and ‘contains the formal proof. To keep the exposition simple, consider just one “that in turn reduces their incentive to make effort. In this case, if a policy measure, namely a reward for undertaking the risky activity. Su | competitive equilibrium exists, it must have zero effort and complete cach of the (V~M) workers in the X sector is given a small lump sum ds “Insurance. To see this, suppose the equilibrium level of cach individual's This induces migration to that sector. As M falls, the Y sector wage ris Hello. is e. With perfect competition, xe)/(1-m(e)) units of low state income which eventually re-equates the utilities from working in the two sectors. E/ean be purchased by giving up each unit of highstate income. If the Aes are further effects. To finance dhe cant, we must raise the isdividual purchases z such contracts, tax on all workers by 4g:(N—M)/N. Furthermore, as the Y sector rises by w, the profits there change downward bydiI<0, and each worker I=pm+e,—t+zale\(1—m(e), as)208 A, Dist, Trade and insurance wth moral hazard Iy=PXat$u-!—2 Then z is chosen to maximize (1=n(@)V(L, p)+ (6) VL?) ~e- ‘This gives the first-order condition: Vale P= Vill?) which implies f= (complete insurance) and V(ft, p)=VUyp). Then choice of effort can only be optimal if e=0. Amott and Stiglitz (1986) this the normal case in absence of exclusivity Suppose the economy is in such an equilibrium. What role can our policy instruments p, g, and gy play? None of them can alter the zero level of effor or the completeness of insurance: those properties were derived above for: arbitrary p, g, and gy. Thus, we have in effect an economy with effort fixed at zero, a fixed accident probability x(0), no moral hazard, and | insurance. Standard theorems ensure that the competitive equilibrium | Pareto optimal (conditional on the effort level), and the policies can do 1 better. Interested readers can derive the result formally by methods simil those of section 2. ‘What this situation needs is a different kind of policy ~ a tax at a suitat chosen rate t on the purchase of insurance. Then giving up a unit of high state income will get you only (1—z)a(e}(1—x(e)) units of low-state income People will purchase incomplete insurance, and make some effort. See Arnk and Stiglitz (1986) for a full discussion. Of course to tax insurance purchas the government must observe it, and then a simpler policy may be to enfor exclusivity. The outcome with zero effort and full insurance may be the most natura ‘one in the absence of exclusivity, but it is not the only one. Moral fi contributes a non-convexity to indifference curves in a state-conti consumption space. In section 2, exclusivity permitted the use of quanti ed insurance, ie. non-linear budget sets, that overcame the problem in the absence of exclusivity, that cannot be done, and a competi equilibrium may not exist. Some new possibilities arise. Stiglitz (1983). discusses them in detail. There may be an equilibrium involving randomize insurance contracts. There may be a non-market-clearing equilibrium with sticky benefit-premium ratio. Finally, there may be a. reactive ot s phisticated coniecture equilibrium with no insurance. Each insurer is ten to offer a contract with a small coverage that will be profitable on its But each realizes that when others act similarly, the resulting coverage § be much greater, the effort level will be much lower, the probability off A. Dist, Prade and insurance with moral hazard 29 bad state will be much higher, and the profit on his contract will become | egative. This last possibility is not a Nash equilibrium, and therefore its intellectual foundation is shaky. But it is the only case where moral hazard is the cause ‘of a complete cessation of private insurance, which was the staring point of | the ad hoc incomplete-markets models of Newbery and Stiglitz (1984) and Eaton and Grossman (1985). Therefore I shall examine it in some detail to see how trade and tax-transfer policies by themselves can cope with moral hazard. ‘The model is the same as that of section 2, except of course 2, and 24 are | fixed ar zero, Ax and x are irrelevant, and correspondingly we drop the four ‘equations (7) and (9)-(1). With private markets absent, some government insurance (gy.y) can be provided on an exclusive basis. If private insurance remains absent or can be prohibited, the government can trivially implement the constrained Pareto- ‘optimal level of public insurance. More interesting is the role of trade policy. ‘Once again, I shall offer an informal reasoning here, relegating the details to | appendix B. Suppose the domestic price is changed by dp. Since an X-sector worker | who realizes the high state produces xy and consumes cy, his real income rises by (xy —cq9 4p. Similarly, we have (x,~c,) dp for an unlucky X-sector worker. The offsetting effect on the per capita tax for everyone is a weighted __ average of the two, say Yala Cw) SP +7, er) AP. | where 74 and 7, are positive and sum to one, |The purpose of insurance is to shift some income from the high state to _ the low state. Therefore we want (6) 4p Diuln— en) 4? + 7%, — 44) AP] >O “and (ned 4 Drude) 40 +704, =x) AP] <0. ‘These simplily to | [oi —ex) (ane) 4p> 0. a) To see what (21) implies for trade policy, note the budget constraint for “each state i=L,H: poids mint219 A. Dixit, Trade ond insurance with moral hazard where d; is the consumption of the Y-good. Then (x,~¢) becomes: (4,9 4p>0. Ap<0. The domestic price of the risky good should be kept below its world price, Therefore in this model, trade policy provides insurance by being counter-protective. This may sound paradoxical. The poiat is that the policy is in the fast instance taking away more zeal income from those who realize. high output than from those who get low output. The average is thea returned to cveryone by means of the lump-sum transfer. The net result is that those in the bad state are net gainers and those in the good state are net losers; this is how insurance is generated Other means of collecting or disbursing revenues will have diferent consequences. In a similar context, but without moral hazard, Grossman (1984) uses a proportional income tax and finds that free trade is the constrained optimum, provided preferences are homothetic. Now moral hazard has no first-order effect on utility because of the Envelope Theorem, and on government revenue because the initial point has (p—p*), &, and gy. all zero. Therefore Grossman's result would continue to hold if proportion taxes were used in my model If the marginal sbift of the tax schedule that is used for collecting oF disturbing incremental revenues is progressive (a feature that can be quite Independent of the prugiesivity of the schedule for fixed revenus require ments) then a small positive tariff will be beneficial. My conclusion from all. this is that the role of tarifs as second-best insurance schemes, when risk sharing markets shut down because of moral hazard, is ambiguous at best, and depends on an accidental conjunction between trade policy and income taxation, ft is much simpler to find and implement a welfare-improving. public insurance scheme. 4. Aggregate risk, exclusive insurance In this section 1 reimpose the exclusivity requirement on insurance, bul introduce tome systematic or aggregate risk. It would not make serie remove individual risk entirely. For suppose one person’s output is function of his own effort e and an economy-wide random variable 6. 1 relative outputs across people convey precise information about their relat effort levels. The full-information first best can be achieved, or clos approximated, by payments schemes based on relative performance. If mi hazard is to have some role, some individual risk is essential. | Tor clai A. Disit, Trade and insurance with moral hazord aun I shall therefore suppose that for each X-sector worker the probability x of high output is a function of his effort e and an economy-wide random variable 8, and that for cach given @, these events arc independent across workers. Let 9(6) be the density function of 6. (There is a fully equivalent ‘model with discrete values 6, and probabilities 6,. Also, a slight change of notation allows us to treat the world price p* itself as @ systematic random variable; the results below are unaffected.) Even in this setting, relative performance schemes have some scope. In my model, it is possible to do even better by conditioning everything on 0, as we shall see in a moment. Hart (1983) constructs a model where a competitive ‘market itself works like a relative performance scheme. That cannot happen here because the economy is small and p is fixed. Remember that my ground-rule of faimess requires that we give the ‘markets for risk-sharing their fall role, Now the large numbers assumption ensures that, conditional on @ the output per worker in the X-sector is riskless and is given by (0) =(1—nte, 6))x, + me, A. 22) | Since the equilibrium value of e can be computed, and output can be ‘observed ex post, we can infer 0 ex post. Therefore all ex ante contracts and policies can be conditioned on it. So we can have a trade policy defined by a fonction p(6), and transfer or insurance policies g,(6), 40), with the associated lump-sum tax (6). Exclusive private insurance contracts can be similar functions (6), (0) People can also trade contingent claims to purchasing power, or Arrow’s securities. I shall write q(@)¢(0) for the price of a claim to a unit of the numéraire if @ occurs. This is a harmless renaming that simplifies the notation by making budget constraints expectations with respect to @. The equilibrium profile q(@) can be found from the conditions of market-clearing 3s for each @; fortunately we do not need the explicit solutions. A new feature is that with p and t dopendent on 8, the Y-sector workers are also exposed to risk. They will therefore participate in the contingent dlaims market. Suppose their purchases are by(0). Then JyO)= w+ by(0)— 40) 23) Suppose the wage w, although sure, is paid ex post. (The altemative assumption is a mere algebraic reformulation with identical results) Then the budget constraint while trading in contingent claims is J4{0}4(0)by(0)d9=0.12 A. Dist, Trade and insurance with moral hazard A. Dist, Trade and insurance with mora hazard 213 I shall write this more compactly as The condition for by(6) is just the sum of these two and therefore redundant; this is obvious since the insurer could always dispense with by(0) as such and just add it on to each of (0), 24(8). The condition for e is E{urtede, OVO), (9) — VO), A) + dre, )[2(6)—2y()]} =0- G4) EoLa(A)by(6)] =0, where E, denotes expectation with respect to 8. The maximand is ty = El ViTy(6), (8). This problem has the first-order conditions: |The remaining equilibrium conditions are: the requirement of indifference in occupation choice, Vl), MOD = 2G), Ux = ys 35) for all @, With no moral hazard in this sector, we have first-best risk 9 allocation, : In the X-sector, we can likewise write: and the state-by-state government budget balance condition, Nt(@)= —(F(M)—MF(M)] 18) = PBX, + 86) + 218) + Bx 8) ~ 19), +(N—M)}[(1 —a(e, O)g.(0) + ale, gy(8): 140) =H Ory + su(0)+2H40)+b9(0)—H), ae and — (PE) — PLN — MYL nhc, N(x) — ug = al —n(e, 99) VU). + mle, (0), (8) —€ + nle,6)(cH(@)—xn)] + MeO]. (36) Begin with the optimum choice of e, given all the other things. The fi order condition is Policy changes in this model consist of shifis of the whole functions p(0), tc. Although this is an added degree of complexity, the structure of the model is very similar to the case of individual risk in section 2. Therefore it is no surprise that trade and tax-transfer policies are once again incapable of improving upon the market equilibrium. Even with systematic risk and moral hazard, competitive and exclusive insurance and state-contingent just as do z,(6) and zy(6). Therefore we should apply exclusivity to all thr claims achieve a constrained Pareto optimum. The details are in appendix C; thus requiring each X-sector worker to gct all his insurance for aggregate the intuition is similar to that developed in section 2. and individual risk from the same insurer, [An equivalent model can be had T hope that cases with aggregate risk and non-exclusive insurance may if bx(6) must be committed before 2.(@) and 24(6) are determined.) now be left to the reader. If equilibrium has zero effort, full insurance for the The insurer's zero-profit constraint is individual risk and firstcbest allocation of aggregate risk, then trade and transfer policies are powerless. If the market shuts down, trade policies have the same kind of complex role, and publicly provided insurance a simpler role, as in section 3. Eolz cle, OVW), A) — VEE), OT = 1, Eq{{(1— nie, 0)z(0) + x10, 6)z(A) + bx(A)]Q(0)}=0. He chooses 2(6), 24(8) and bx(6) for all @ to maximize (29) subject to (3 and (31). The first-order conditions for z,(6) and (0), respectively, are 5. Concluding comments (1 —nfe, 8) — axle, HVAT), (O)) — Axl! ~ mle, O)q(G)=0, ‘These results should not create the impression that moral hazard never matters, We know from the works of Stiglitz and his co-authors cited (ne, 9) + un Ce, Oil lu), (9) — Arte, O)q(6)=0. throughout the paper that it does. There are important exterhalities and non-Vibe tr tea ea silk er here A. Dixit, Trade and insurance with moral hazard 21s convexities associated with moral hazard, Complicating my model will introduce some of them. (1) If there are several risky activities with moral hazard, there can externalities through their interaction, although such effects can be inter: nalized by extending the exclusivity requirement. (2) If the economy is large, producer prices in free trade are endogenous. Changes in them affect producers’ profits which become consumers’ incomes, this alters the incentives 10 make efforts to reduce the risk, The standatd monopoly optimum tariff can handle that. () TT there is distorting commodity taxation for come other reason such as income redistribution or the provision of public goods, then there can be similar income effects via the tax revenue, : (4) If an individual's probability of success is a function not only of his unobservable effort, but also of his anonymous purchase of a commodity, then welfare could be increased by subsidizing the consumption of th commodity. However, this is not a case for tariff, ie. for taxing domestic production and imports of the commodity differently. a (5) I have not considered non-market-clearing equilibria. These and other problems will enlarge the scope for beneficial policy interventions, although I _ believe the role of trade policy will remain complex and ambiguous. a All these limitations are significant, and indicate the need for more research. But I think the paper points out even more serious limitations of previous work on trade policy under uncertainty. Using models very similar or identical to those of the carlier authors, but taking explicit account of o of the causes of incomplete risk-sharing markete on which they tacitly rely, 1 have found that their ad hoc assumption of missing insurance markets does. indeed drive their activist policy conclusions. This is yet another instance of the need for a ‘level playing field’ when economists compare the performance of markets and governments. ® defined in (8), ie. the effect holding the choice variables and the ‘multipliers unchanged, and then evaluate the result at the optimum. There- fore we have Aux= (1 —nfe)— pre DEVAT PY. 4p + den — At) + Ville.) dp) +(e) + ure(e)) Vill us P) (Xn AP + 48y— 4t}+ Vill, P) Ap] = (tafe) — ur (Vl PL 44) 49+ 48, ~ 407 +(e) + ure) Milly P)Cn— ew) Ap + Agu At] =Ax{(1 ne) [hy 01) 4p + 4g) + H( ney) A+ Ay) — At}. an The first step follows from the Envelope Theorem, the second from Roy's Identity, and the third from the first-order conditions (9) and (10). Note that if we conduct the thought-experiment of giving a sure dollar to an X-sector worker, ie, formally set dgy=dg,=1 and Ap=Ar=0, we get Au roves that dy is the marginal utility of sure income in the X-sector. Turning to the Y-sector, we have uy = Villy. p)(4w~ 49+ V(ly,p) Ap =A FM) AM — At—ey Ap) (a2) Now we can set Auy=duy. Define K as 2y/y, the ratio of marginal utilities in the two sectors, Then we find Appendix A FM) AM =r +ey 4p+K((1 ate) [xy ex) 4p +. 4e,] risk and exclusive insurance. The subject is the formal analysis of the eects of the policies represented by p, a, and gy on the equilibrium. Starting fom aii initial position where none of these polices is used, ie. p=p* and g,=gy—0, consider a small change 4p, 4g, and dgy. This will in general change all the magnitudes, eg. M by AM, ¢ by t, ete : First, focus on the problem of devising the equilibrium insurance contract, This takes p, gi, gy and t as parametric, and chooses =, zy and e to maximize ux as given by (5), subject to the constraints (6) and (7). When the + nfe)[xn—c) AP + 48q] — 41}. as) Finally, differentiating the government's budget constraint around the initial point of g, p= p*, we have N dt= MF"(M) AM +(N—M)[(—n(e) Ag, +2(2) Agu) —p{(N—MYL[(1 —nfe)(e, x1) + (ecu u)] + Mey}. (AA)216 A. Dixit, Trade and insurance with morat hazard A. Dist, Trade and insurance with moral hazard Substituting for F°(M) AM from (A.3) and simplifying yields: F'(M) AM =At+cy 4p +[Ky(1— me) (x, 21) + Kumte) mn — end] 4p Atm(1— nie) dg, + m0) AB + 0K (1 —nle) de, + Kure) det] +11 mer ed) + 212) u— ew) 4P- ‘Then (A.1) becomes 4uy=0, and so of course duy=0. Workers’ utilities (7 5 are stationary with respect to small policy changes around the initial passive position, ~[K,(1 16) + Kyte] 4t (N= M4MK,(1— (0) + MKyr(e)] At Appendix B Here we have individual risk, but insurance is not constrained exclusivity. Begin with the passive policy and consider small changes. =[0—m(e)(N-M4+MK oy —e,) +n(e)(N—M + MK (xq —en)] 47 to maximize uy (the Envelope Theorem again), we have 4+. —x(e)\(N-M+MK,) 4g, Aux = (= 210) Vi Pe. AP + Ag — 40) + VCP) API +n(e\(N-M+MKy) dee 4+ nfe) Vln B)(u A+ dy 40+ Vln) AP) oe This can be written in a more transparent form as =(1—nfel)A Loa.) 4p + 4g, — 44] At=y.[,—e1) 4p + 481) + yall ex) 4p + 4g), (B.7) +n(@Viulxu—eu) AP + Agu 40), | where the weights y, and yy are readily defined by reference to (B.6), and add where I have defined A=VAlLP n= Villu) : : up to unity. Finally, combining (B.1) and (B.6), we find that, leaving out some positive constants, Aux ~ (iy — Ay) {Loe = Cea ewd] AP + 48, — 48} (BS) With A, >Zyy we see that utility can be increased by suitable policies that start from the passive stance of p=p* and g,=#4=0. The general principle __ is that it is desirable to provide some insurance despite the moral hazard. Thus, tax-transfer policies that have 4g,>0 and/or 4gy<0 can increase ‘welfare. Note that if the latter is used on its own, we have dt=7u den<0, 30 wwe have 4, > Ay. The marginal utility of sure income is A 1 —m(e))2, + me) For the Y-sector workers, we have (A.2), namely uty = Ay FM) AM —At—cy Ap) Ba) as before. Now we set dux=duy and simplify. Writing Ky=A,/ay and dy, we find ay That is how income is transferred in the desirable direction by such a policy. Ky218 A. Dist, Trade and insurance with moral hazerd ane naieaa dara tiie a What about the role of tariffs as insurance? The desirable direction Py Lond in: ie change, from (B.), is to make ‘by L and H as appropriate. Then similar steps give Aug =By{ 1 2 1n EV Le. es [.-4) (aye) Ae By{[1 aye) (Vi) rs 4p + Age — 40) + VL) 4p) (e+ ure) VHD) xy AP + Aga 40) + VE) spy} incomes, namely (fy —e, 4p) and (414—cq 4p). When trade policy is on its , =e = = ‘own, using (B.7), we have . | BAC esters eh aet Ae) + un PAE Uy _ Aly ey p= yullx, —c1)—(%n—cw)] Ap, ee eee eee ix Fo{( —)al (x, 01) 4p + 4g, — 4 aio ipa (1 —m)al, 41) 4p + 4g, — At] valle) nen). +nal(x,—er) p+ Ag, — 4t]} ‘Therefore income is shifted toward the worse state when (B9) is satisfied. ix Eo(af(l 7) (1 —e4) + (en —cw)] dp pendix C : +All 7) Ag, +7 Agu)—a 41}. (C2) Here we have some aggregate risk, but exclusive insurance, Start with: As before, define K=Ay/dy, set Auy=duy, and ri ‘This passive trade and transfer policies, ie. (0)=p* and g,(@)=g,40)=0 for all yields: Deere rere eeeeeee een and consider a small change 4p(@), dgy(9) and dgy(@). By these I mean shifts a of the whole functions p(9}, ete. and not movements along them such a FM) 4M=Eyla 40] —Faley 4p) (6) 40. Calculation of the effects on utilities proceeds as in appendix A, com: plicaed only slightly by the need (© take expectations with respect € 9 Thus, +KEy{Ql(l—i(x1 — cr) + 20xu—cu) Av +al(1—2) 4g, + dgu]—a 40} (C3 ty = Eg Villy(@), (ON) Aw — At 0)] + Vel y(6), 2(8)) 460) [Differentiation of (36) for fixed 6 gives = Bo{ Vill y(6), (0)) [Aw — At(8}—e(8) 4p49)]} N 4t=_MF"(M) AM +(N~M)[U—x) 4g; +2 dgu] = Ayal (ONL AW— 0) (9) 4040)]} —4pf(N~M)[(1 =n)(6, 4) + nlc — mu) + MeyT =Ay{dw—E,Lq(8) 44(6)) —ELey(8)q(9) 4p(@)]}- _ where once again I have omitted the arguments @ or (e,8). On multiplying by | 0) and taking expectations, ‘The first step uses the Envelope Theorem, the second, Roy’s Identity, and third, the first-order condition (25). The last step uses the fact that Byg( being the price of a claim that pays a unit of the numéraire in every sta rust equal one. The expression for du, gets far too lengthy. Therefore I shall omit the arguments 8, and indicate the points where V; and V, are evaluated simpy NE(q 4] =MF(M) 4M +(N~M)E,{q((1 —2) 4g, +7 4gy)] +(N~M)EeLgl(l ~2) (x, -c1) + 21xu—Cw)] 4p] —M Eley 4p], cyA Dhuit, Trade and insurance with moral hazard Lal —2)(4, ex) + 0(%%— ew) AP) + EsLal(l —*) 4g, +74 g4:J]- When we use (C.5) in expression (C.2) we find Ay = and Joseph Stiga, 1986, The welfare economics of moral hazard, Quee’ ‘University Disevsion Paper no, 635. Diamond, Peter and James Mirtlees, 1971, Optimal taxation and public production 1 ‘American Economic Review 61 (1) Marchand (2) June, 827, 261-278 Exton, Jonathan ané Gene Grossman, 1985, Tarifs as insurance: Optimal commercial poiay ea dome marke nen, Canadian ara of Eee 8 oh ag ST Greenwald, Broce and Joseph Stig, 1986, Extemaities in economies with imperfect fale mation and incomplete markets, Quarterly Journal of Ezonomics 10! (2, May, 229-264, Grossman, Geng, 1984 International trade, foreign investment and the formation of he ‘entrepreneurial class American Eeonomic Review 14 (9, Sepe, 605-54 Hart, Oliver, 1983, The market mechanism a2 an incentive scheme, Bel Journal of Economics 14 (G), Autumn, 366-382 Han, Oliver and Bengt Holsrom, 1986, The theory of contracts in: Truman Bewley, ‘Advanoes in esonomie theory, 1985 (Cambridge University Pres) fortooming : Nevo, Davin ep St 184 Pare ior de, Revi of Eeanomi Stu Stiglte, Joseph, 1983, Risk, incentives and insurance: The pure theory of moral hazard, fnnurance 8, no. 26, Jenn 4-93. Journal of Inteentional Economies 23 (1987) 221-239. North-Holland BORROWING TO DEFEND THE EXCHANGE RATE AND THE TIMING AND MAGNITUDE OF SPECULATIVE ATTACKS Willem H, BUITER* Yale Universiy, New Haven, CT 08320, 05.4 (NBER and CEPR Received March 1986, revised version received Janvary 1987 |The paper analyses the implications for the timing and magnitude of « speculative attack on a {ned exchange rate of goveenment borrowing to increase the stock of intermational reserves. Iti F shown that such borrowing reduces the likelhood of an catly collapse and increases the | telinood of an eventual collapse, The magnitude ef the speculative atiack also increases. 1. Introduction The purpose of this paper is to extend the recent literature on collapsing managed exchange rate regimes by allowing explicitly for the government | budget constraint and especially for the interest cost of servicing the public __ implications of government attempts to peg the price of gold or to defend a price ceiling by managing a stockpile; it demonstrated the inevitability of an eventual speculative attack and of the collapse of such schemes. Krugman (1979) provided the first application of this analytical approach to the there have been many other, mainly theoretical, studies of the viability of a | Nariety of managed exchange rate regimes and of the nature and likelihood ‘of speculative attacks. (See, for example, Flood and Garber (1983, 1984a, 1984b), Obstfeld (1984, 1986a, 19866), Connolly and Taylor (1984) and Dornbusch (1984)]. Related relevant theoretical work on speculative attacks, runs, etc, can be found in Salant (1983) which analyses price stabilization schemes and Diamond and Dybvig (1983) which deals with bank runs. - Empirical work in this tradition is still relatively scarce, but notable examples | are Cumby and yan Wijnbergen (1983), Grilli (1984, 1986b), Collins (1984) "and Garber and Grilli (1986) *L would ike to thank Sweder van Wijnbergen for many useful discussions on the subject of | seal policy and tae viability of managed exchange rate regimes. An snonymous releree made ‘many helpful suggestions