Capital-Account Liberalization as a Signal

by Allan Drazen, Leonardo Bartolini
Citation
Title:
Capital-Account Liberalization as a Signal
Author:
Allan Drazen, Leonardo Bartolini
Year: 
1997
Publication: 
The American Economic Review
Volume: 
87
Issue: 
1
Start Page: 
138
End Page: 
154
Publisher: 
Language: 
English
URL: 
Select license: 
Select License
DOI: 
PMID: 
ISSN: 
Abstract:

Capital-Account Liberalization as a Signal

We present a model in which a government's current capital-control policy sig- nals future policies. Controls on capital outflows evolve in response to news on technology, conditional on government attitudes toward taxation of capital. When there is uncertainty over government types, a policy of liberal capital outflows sends a favorable signal that may trigger a capital injlow. This prediction is consistent with the experience of several countries that have liberalized their capital accounts. (JELF21, C73)

Controls on international capital flows are a lead to a capital outflow, as funds flow to common form of financial regulation. Changes where returns are highest (see e.g., Maurice in the extent of these controls are also com- Obstfeld, 1984; Philippe Bacchetta, 1992), mon. Of 182 countries surveyed by the Inter- while removal of nonbinding controls should national Monetary Fund (IMF) in 1995, 129 have no effect. were classified as restricting international cap- Actual experience with lifting controls, how- ital transactions. At least 50 of them had sig- ever, tells a different story. Many countries that nificantly altered these regulations in the have removed controls on outflows have exDe- previous 12 months. rienced rapid and massive inflows of capitali A

Despite the widespread use of capital con- popular explanation, motivated by the work of trols and the frequency with which such re- Michael Dooley and Peter Isard (1980), was strictions are modified. little work has been formalized by Lab& and Larrain (1993). Condone to model the impact of changes in these trols that prevent investors from withdrawing regulations on capital flows. Perhaps this de- capital from a country act like investment irre- ficiency reflects the view that the effect of lift- versibilitv. Their removal makes investors more ing or imposing controls seems clear from willing to invest in a country, as it is easier to basic economic theory. Consider, for example, get their capital out in the future. the removal of restrictions on capital outflows. However, the link between capital controls If controls are binding when the liberalization and investment flexibility, though essential to is implemented (so that offshore returns ex- any model of capital controls, provides only a ceed onshore returns), a liberalization should partial explanation of capital inflows. This

explanation depends crucially on expected

persistence of current policies, but unlike

technological constraints, policies may change.

* Bartolini: Research Department, Federal Reserve Bank of New York, 33 Liberty Street, New York, NY In fact, governments that succeed in attract- 10045; Drazen: Center for International Economics, 4 1 18 ing foreign investment have a strong incen- Tydings Hall, University of Maryland, College Park, MD

tive to lock the door once the fattened calves

20742. We thank M. Annunziata, M. Baxter, G. Calvo, G.

have come inside. To make sense of the flex-

M. Milesi-Ferretti, A. Razin, N. Roubini, A! Tornell, two referees, and participants in seminars at Boston College, ibility argument and motivate the persistence the European University Institute, Georgetown University, of policies affecting capital mobility, one Harvard University, Hebrew University of Jerusalem, the needs a model that captures the interaction International Monetary Fund, the University of Maryland,

between optimizing, forward-looking inves-

the Massachusetts Institute of Technology, the National

tors and governments.

Bureau of Economic Research, New York University, the Federal Reserve Bank of New York, Penn State, UCLA, the University of Wisconsin, and the World Bank for com- ments and suggestions, and N. Gunaratne and A. Peterson ' Francesco Giavazzi and Luigi Spaventa ( 1990), John for technical support. This paper need not reflect the view Williamson ( 1991), Donald Mathieson and Liliana Rojas- of the Federal Reserve System. Errors and omissions are Suarez (1993), and Raul Labfin and Felipe Larrain our responsibility. ( 1993), among others, review a number of such episodes.

138

Our approach to modeling capital controls and explaining the observed inflows following the adoption of a regime of liberal outflows views capital controls as potential signals of future government behavior. Specifically, we suggest that, besides providing greater flexi- bility for current allocation of capital, a regime of free capital mobility may signal that impo- sition of controls is less likely to occur in the future and, more generally, that future policies are likely to be more favorable to investment. Our argument rests on the belief that investors have imperfect information on governments' intentions and constraints and, therefore, may use the observation of current policies toward investment to infer the course of future poli- cies. This gives governments an incentive to allow free capital mobility so as to provide a favorable signal on future investment policies. If the signal is successful, capital flows in.

A signaling model must make this motive consistent with the ultimate purpose of capital controls, which is often to broaden the domes- tic tax base. Countries with poorly developed tax systems often rely on revenues from finan- cial repression, enforcing a differential be- tween onshore and offshore returns to capital by regulations aimed at "trapping" capital on- shore. In our model, it is precisely those gov- ernments that depend most on such a tax base that impose controls, even though such con- trols may lead to a lower expected tax base.

To make the argument convincing, we must show that governments use capital controls as an equilibrium response to information they receive: the choice of an open capital account signals good news about the future, and vice versa. However, if adoption of a regime of free capital mobility is expected to lead to a capital inflow, why would a government that expects bad times not attempt to take advantage of this? (Formally, the question is: why does a separating equilibrium prevail, rather than a pooling equilibrium in which all governments choose not to impose controls?) To answer this question, we show that some governments choose to impose capital controls, even though they know this is interpreted as an unfavorable signal.

The argument behind this result is simple. Consider a government that raises revenue from several sources, including capital taxa- tion, to finance the provision of public goods.

Suppose that revenues (and hence expendi- tures) have a stochastic component and that government welfare is highly concave in the level of expenditure, so that low expenditure implies very low welfare. A government that anticipates low revenues from other sources is especially sensitive to the possibility of low capital tax collection. It will then impose con- trols to self-insure against bad states of nature (when capital would flow out), thereby assur- ing a minimum level of revenues in all states of nature. It will impose controls, even though by doing so it may forgo higher revenues on average across all states.

Our approach has both strengths and limi- tations. First, in contrast to most previous stud- ies, we model capital controls as a dynamic component of governments' problem, rather than as exogenous constraints applying over the whole horizon. Second, unlike the standard symmetric modeling of controls on inflows and outflows, we recognize that real-world controls are typically asymmetric: stricter ei- ther on inflows or (more frequently) on out- flows. Greater realism in these respects comes at the cost of simplification elsewhere. For in- stance, we do not distinguish between restric- tions on short-term (or portfolio) flows and those on long-term (or foreign direct invest- ment [FDI]) flows, nor between restrictions on residents and nonresidents-although one may imagine situations where liberalizing res- idents' portfolio activity alone may send a fa- vorable signal, thereby indirectly promoting other inflows. Also, we focus solely on poli- cies affecting capital mobility, although we are aware that capital-account liberalization is of- ten only one element of broader reform pro- grams. However, by focusing on a single motive for inflows, we can make the clearest case for the signaling role of policies affecting capital mobility. Our model suggests that im- perfect information about a government's in- tentions may provide an incentive to use free capital mobility to enhance the credibility of a broader reform program. It is to a reformist government's advantage to show an early commitment to an open capital account, by ex- posing itself to risks that less-committed gov- ernments cannot afford.

The paper is structured as follows. The next section reviews the liberalization experience of a number of countries, pointing to stylized facts that are consistent with our model of cap- ital controls but may be more difficult to ex- plain by alternative models. The model is described in Section I1 and solved in Section

111. Section IV discusses the results, and Sec- tion V concludes.

I. Capital Controls: Some Stylized Facts and Liberalizations Episodes

We begin with a few stylized facts. First, capital controls are much more common among developing countries than among in- dustrial countries. At the beginning of 1995, for instance, capital controls were used by 126 of 158 developing countries, but in only three of 24 OECD countries (Greece, Norway, and Turkey ).

Second, capital controls are predominantly aimed at restricting capital outflows. These controls take a variety of forms, including quantitative restrictions and outright prohibi- tion of outflows, requirements to surrender a portion of the outflow to a low-interest-rate ac- count, and dual exchange rates. Though these regulations look different, in practice all aim at stemming outflows by making their cost prohibitive.

Third, capital controls appear to play two main roles: either in support of governments' attempt to broaden the tax base for a capital levy, inflation tax, and various forms of "fi- nancial repression"; or in support of fixed- or managed-exchange-rate policies.

Finally, liberalizations of capital outflows are often accompanied by a sharp increase in net capital inflows, as the experience of some countries that have recently liberalized their capital accounts illustrates (see Figure 1 for summary data). The four episodes we review are those of Italy, New Zealand, Uruguay, and Spain.

Italy began to dismantle its system of con- trols on capital outflows in November 1984. The compulsory zero-interest deposit on port- folio investment abroad was reduced for resi- dents and abolished for mutual funds. The surrender requirement was further reduced in 1985 and 1986 and abolished in 1987, and the crediting of banknotes to capital accounts was liberalized in August 1986. Though the liber- alization was completed only in May 1990, its main steps occurred in 1986 and 1987, after which remaining restrictions ceased to be binding (see Bartolini and Gordon Bodnar [I9921 and Giavazzi and Alberto Giovannini [I9891 for analysis of this episode). Large inflows were recorded from 1987, and pri- vate investors were primarily responsible for them.2

In November 1984, New Zealand abolished the exchange and capital controls that had been in place since 1938, as part of a broad policy of financial liberalization. In contrast with the policies followed by Italy and Spain during the 1980's (where the liberalization was part of a policy of greater exchange-rate fixity), New Zealand floated its exchange rate soon after the liberalization. The capital lib- eralization was rapid and focused on the lifting of restrictions on outflows, including the sur- render of foreign-exchange receipts and limits on holdings of foreign securities, and on rais- ing of domestic funds by foreign companies. Interestingly, although the financial market was liberalized in June and July 1984, capital inflows did not surge until year-end, when the capital account was liberalized. In fact, the net inflow recorded in 1984 appears to have oc- curred wholly in the last two months of the year. Capital inflows surged in 1985, 1986, and 1987, with private investors playing, once again, the principal role.'

After nearly two decades of inward-loolung policies and financial repression, Uruguay began to implement radical reforms in the mid-1970's, including trade and financial de- regulation, and-foremost-liberalization of capital flows.4 The liberalization of capital flows proceeded at the fastest pace and before

'As documented in Figure 1, (net) capital flows minus (net) loans to the public sector surged beginning in 1986- 1987. (Although Italy does not classify capital flows by characteristics of foreign investors, flows from official agencies and other governments should be almost entirely included among foreign loans to the public sector.)

'Although the hands-off approach following the lib- eralization limits data availability, net flows to the private sector were noted as growing faster, and to a higher level, than those to the public sector (see Reserve Bank of New Zealand, 1986a, b; OECD, 1987). Even flows to the pub- lic sector originated wholly at market terms, mainly through the sale of government stocks and Kiwi Bonds.

See Ariel Banda and Michele Santo ( 1983), Jaime de Melo (1985), and Juan PBrez-Campanero and Alfredo Leone (1991) for reviews of the Uruguay episode.

Percent of GDP Millions of US$ Percent of GDP Mtlllons of US$

,-3.0

I

I ,70000

Percent of GDP Milltonsof US$1 350
-Capital account (net of FDI for Spain); left axis. Foreign-exchange reserves; right axis. ---Capital account minus net loans (net official loans for Spain); left axis ---  

Notes: Shaded areas indicate periods of main liberalization measures. Capital flows include errors and omissions.

Sources: IMF, International Financial Statistics; Banco Central del Uruguay, Indicadores de la Actividad Economico- Financiera; Banca d'Italia, Rapporto Annuale; Banco de Espafia, Boletin Estadistico; The World Bank, World Debt Tables; and data provided by the New Zealand Department of Finance.

other major policy changes: in September banks for the first time, and repatriation of cap- 1974, exchange-rate controls (primarily on ital and profits connected with FDI was per- outflows) were eliminated, residents were per- mitted. Large capital inflows were recorded mitted to hold dollar accounts with domestic in the four years beginning in 1974. Private investors (mainly from Argentina and the United States) played an essential role also in this episode, a feature we document in Figure 1 by netting official net loans from net capital inflows (assuming negligible official acquisi- tions of equity and real estate, and negligible foreign investment in the domestic public sector).

After its entry into the European Community in 1986, Spain liberalized capital flows, as part of a broader plan of fiscal and monetary reform. Both capital outflows and inflows were liberal- i~ed.~

However, evidence from offshore- onshore interest differentials shows that, prior to the liberalization, controls on outflows were more stringent than on inflows (JosC Viiials, 1990). In fact, after the liberalization, the positive offshore-onshore interest differential disappeared and then turned negative when temporary controls on inflows were introduced to stem the rapid inflow recorded in 1987. Ex- cluding measures affecting FDI, the main steps in the liberalization included the lifting of re- strictions on residents' direct and portfolio investment abroad, on forward exchange op- erations, and on real estate investment abroad. Although the liberalization was completed only in 1992, net capital inflows (net of FDI) surged immediately after the initial steps of 1986 and continued unabated until the crisis of 1992. Once again, private investors played a primary role in the inflow, as documented in Figure I, which nets changes in the stock of official loans from total capital flows (exclud- ing FDI) (see Susan Schadler et al. [I9931 for a discussion of this episode).

These episodes can be summarized as fol- lows: the liberalizations focused on removing restrictions on capital outflows; they repre- sented early ingredients of broad reforms that included the lifting of various elem,ents of financial repression; and they were accompa- nied by a surge in net capital inflows. In the next section we present a model capturing sev- eral of these stylized facts.

'Liberalization of inflows affected mainly FDI in se- lected sectors; hence, to avoid overstating our case, since the FDI growth may reflect the lifting of these restrictions, Figure 1 reports data net of FDI.

11. The Model

Consistent with the arguments outlined in the introduction, in our model decisions on capital controls are driven by governments' desire to increase the stock of domestic capital. Our main results require government behavior to display two basic features: first, government utility (net of the cost incurred when imposing capital controls) should increase in the stock of domestic capital; second, the willingness to incur the costs of controls should differ across governments in a way (at least, partially) un- known to investors.

Many of the motives for (and costs of) cap- ital controls suggested in the literature could meet these requirements and fit into our frame- work. For instance, utility from domestic cap- ital could reflect a government's interest in maximizing domestic output, tax base, or of- ficial reserves6 Costs of controls could reflect concern with distortionary effects on capital allocation, penalties enforced by other coun- tries for engaging in beggar-thy-neighbor pol- icies (penalties that may include limits on trade credit, exclusion from participation in coordinated policies, etc.) ,or other economic and political factors that affect the importance governments assign to capital mobility.' Gov- ernments may also be concerned with the ef- fects of controls on residents' ability to smooth consumption through capital flows, although the practical relevance of this motive has been questioned.' A different and important argu- ment concerns the use of controls to insulate

Alberto Alesina et al. (1994) list four main motives for capital controls: (i) limit volatile capital flows; (ii) maintain the domestic tax base; (iii) retain domestic sav- ings; and (iv) sustain structural reform. Alesina et al. then study their determinants across countries and over time. They identify in governments' attempt to collect revenue from financial repression the main motive for controls. (See also Giovannini and Martha de Melo [I9931 and Joshua Aizenman and Pablo Guidotti [1994]).

'Alesina and Guido Tabellini ( 1989), for instance, in- clude the loss of capitalists' electoral support to govern- ments that impose controls among the political costs of capital controls.

Allowing for consumers' risk aversion, for instance, Robert Lucas (1987) and Enrique Mendoza (1991) esti- mate a loss of utility from inability to smooth consumption through international diversification on the order of 0.10 percent. David Backus et al. ( 1992) report similar results.

a country from external shocks. The difficul- ties associated with the recent large inflows and subsequent outflows in Latin America, which culminated in the Mexican crisis of late 1994, underscore the importance of this mo- tive, the analysis of which would require, how- ever, a substantially different model. Although analytical clarity dictates our focus on a single motive for controls, we later discuss how our results may extend beyond our specific model.

To capture these considerations formally, we present the simplest model we could design to yield our main results. We consider a two- period model of a small open economy where a single nonstorable, homogeneous good Y, is produced at time t with onshore capital K,, using a concave technology Y, = p,K?, where p, is a stochastic productivity shock and 0 < 0< 1. The stock of onshore capital K, reflects the decisions of competitive, risk-neutral in- vestors. In each period, investors allocate cap- ital either offshore (at a fixed return r) or onshore, to maximize total expected returns, discounted by the factor p = 1/( 1 + r),over the residual h~rizon.~

(For simplicity we also set the government's discount factor at p.) Ab- sent adjustment costs, the returns to physical and financial capital are equalized, and we make no distinction between the two. For simplicity, and given the ambiguous evidence in support of capital flows for consumption- smoothing purposes (see footnote 8), we ig- nore this motive in our analysis. (Formally, this treatment would follow from the assump- tion of households' linear utility; see Jacob Frenkel and Assaf Razin, [I9871 for a com- plete discussion.) We thus focus on investors' capital allocation decisions in response to technology shocks, given endogenous constraints on capital mobility.'"

The government taxes capital wealth (for simplicity, only at the end of period 2), at a predetermined rate r, collecting fevenues

See Bartoiini and Drazen (1996) for a model (with a different information structure) in which government pol- icies respond to changes in external conditions.

lo As suggested by a referee, our model could also be interpreted as a model of international investment in phys- ical capital, where the multiperiod horizon reflects the ges- tation lag necessary for direct investment to become fully productive.

TK*." Governments differ by the value they attach to this revenue. This heterogeneity may reflect differences in preferences for expendi- tures, in willingness to tap revenues other than capital taxation, or in the importance assigned to free capital mobility-differences which cannot be signaled simply by announcement. We parameterize these differences by a vari- able x E (-w, m), and assume governments to have greater information on x than inves- tors. Although many factors may differentiate governments' willingness to use controls (es- pecially their commitment to nonintervention- ist policies), for concreteness we treat x simply as revenues (or obligations, when x < 0) other than capital taxation, to which the government has access or is willing to use

(or service) at the end of period 2. At the end of period 2, the government transforms the sum rK2 + x into nonmarketed public goods and derives utility W(*) from their supply. This function is increasing, continuous, con- cave, and satisfies the regularity conditions lim,+_,W(z) = -w, lim,,-,W'(z) =m, and lim,,,W '(z) = 0." Thus, government utility from onshore capital is increasing in rK, + x (so that a potentially larger captive tax base tempts governments to impose controls), but at a decreasing rate (so that a larger x reduces the incentives to trap a given stock of capital). We model asymmetric information by assum- ing that governments are informed about x at the beginning of period 1, whereas investors learn this only at the beginning of period 2. We refer to x as a government's type; a higher x identifies governments with greater outside resources or greater willingness to tap those resources. Investors have a prior cumulative probability distribution over types G(x) .

We model the direct costs of controls very simply. (Naturally, "reputational" costs are an integral part of our story and are analyzed below.) Similarly to other models in the sig- naling literature, we interpret the cost of con- trols simply as the cost to the government of

I' The issues that arise when the government is unable to precommit to a fixed tax rate were studied by Stanley Fischer ( 1980). We simplify the model by assuming com- mitment to a fixed tax rate at time zero.

l2Exponential utility would satisfy all these regularity conditions.

TABLE1-TIME STRUCTURE

OF THE MODEL

r = 0: Inherited stock of capital: KO
t = 1: * x is revealed to the government Government chooses cl E {F,R) p, E (0,p) is realized and revealed Public chooses K, (with K, 2KOif c, = R) Production takes place; profits are collected
t = 2: x is revealed to the public Government chooses c2 E(F, R) p2 E (0, p) is realized and revealed Public choosesK2 (with K2 2 K, if c2 = R) Production takes place; profits are collected; taxes are paid and transformed into public goods (together with x)

breaking a commitment to free capital mobility: ''whenever controls are in place, the government pays a cost E > 0. This approach simplifies the exposition by malung the cost of controls independent of whether or not controls turn out to bind ex post. l4

The model's timing is summarized in Table

1. The initial state is summarized by the initial capital stock, KO,and G(x). At the beginning of each period, before observing current productivity, the government announces whether capital flows are free or restricted in period t. When controls are imposed, the end-of-period stock of domestic capital, K,, is constrained to be at least as large as at the beginning of the period, that is, K, 2K, ,,with KO> 0.15(We use a dummy variable c, to denote the period-t regime, letting c, = F and c, = R denote the cases of free and restricted mobility, respectively.) For economy

''See for instance, Kenneth Rogoff ( 1987), where the cost of breaking a no-inflation commitment is modeled as independent of the inflation rate itself, and Robert Barro (1986), where a government's cost is zero for zero inflation and prohibitive for positive inflation.

l4 Little is lost with this assumption, since governments' utility is already a function of Kt andx, and it could be redefined net of other costs that depend on these variables (e.g., as a function of the gap between offshore and onshore returns). Explicit role for residents' welfare could also be included, by defining a government's welfare as utility from public consumption plus consumers' utility from domestic output, produced with onshore capital and a fixed stock of labor.

IF We focus on controls on capital ourflows. As it will become clear later, in this model there are no incentives to restrict inflows.

of exposition, we shall often refer to the adoption of a regime of free capital mobility simply as a "liberalization," although strictly speaking this term should be reserved to describe a switch from a regime with capital controls to one without controls. In Section IV,we discuss when a "liberalization" in the strict sense may indeed emerge endogenously in our model.

After the government has announced the financial regime, nature reveals current-period productivity p,, which may take the values { 0, p} . We initially assume p, to be serially uncorrelated and write the probabilities of p, = 0 and p, = p as 1 -7r and 7r, respectively. We later discuss some implications of the more realistic assumption that p,is serially correlated. After p, has been revealed, investors choose K, (in accord with the announced regime), and profits (as well as taxes, in period 2) are collected at the end of the period.

111. Solution

Solving the model backward from period 2 leads us to a unique perfect Bayesian equilibrium, a standard equilibrium concept in the signaling literature (see e.g., Drew Fudenberg and Jean Tirole, 1991) .

A. Period-2 Equilibrium

After observing p2, investors compare the marginal return from investing offshore (inclusive of principal), p(1 + r), to that from investing onshore, V2= p [p2PKf-' + (1-T)] (inclusive of the scrap value of a unit of capital). With free capital mobility, profit maximization equalizes p( 1 + r) and V2,yielding the optimal level of onshore investment,

which equals zero if p2 = 0.If, instead, capital controls are in place in period 2, investors may be unable to attain this solution, since the domestic stock of capital must satisfy the constraint K2 2 KI. Then,

Thus, depending on the inherited stock Kl, capital controls may or may not bind in period 2 in the high state p, = p. However, controls certainly bind (and the comer solution K, = Kl prevails) in the low state p, = 0, if a posi- tive Kl is inherited from period 1. This possi- bility, given investors' period-1 uncertainty about the risk of controls in period 2, plays a crucial role in the period-1 equilibrium studied below.

At the beginning of period 2, the government decides whether or not to impose controls, given the inherited stock K,. The government's problem can be summarized by the function 41, = 412(KI, x), which defines a type x's expected utility gain from imposing controls. Types for which 412 > 0 impose con- trols in this period, while the remaining types allow free capital mobility (with no loss of generality, indifferent types allow free mobil- ity). We define 412 as

where the indicator function I' = I(K, > K; (p)) equals 1 when Kl exceeds the optimal period-2 stock in the high-productivity state p, = p, and zero otherwise.

Under the model's assumptions [in partic- ular, the concavity of W(*)], 412 = 412(KI, x) decreases monotonically in x from m,to -p E. Hence, the period-2 equilibrium features a low range of types (those for which 412 > O), which impose capital controls in period 2, and a higher range of types with sufficient outside resources such that412 5 0, which allow free capital mobility. This property is intuitive: capital controls raise expected tax revenues in period 2 (the tax base is higher with binding controls and unchanged otherwise), thus rais- ing a government's expected utility from public expenditure. The concavity of W(*) implies that this utility gain falls with x, though. For sufficiently large x, the gain from broadening the tax base is outweighed by the cost of controls. Also, 4, rises with the inher- ited stock of capital Kl : a higher Kl provides a potentially larger captive tax base and, hence, stronger temptation to impose controls. Based on these properties, we now study the equilibrium prevailing in period 1.

B. Period I : Signaling Equilibrium

In period 1, investors also compare the ex- pected returns from onshore and offshore in- vestment. In so doing, however, they must consider the probability that capital controls may be imposed in period 2, a probability that reflects their current beliefs over government types, conditional on the policy chosen by the government at the beginning of period 1.

To study this problem, denote the probability that controls may be imposed in period 2, after having been imposed in period 1, as yR =yR(Ko,K1)=Pr(c2=RIcI=R)andtheprobability that controls may be imposed in period 2, after not having been imposed in period 1, as yF= yF(K0, K,) =Pr(c2 =RIc, =F).(These probabilities depend on both KO and K, ,as these affect the incentives of governments to impose controls in periods 1 and 2, respectively.)

Next, the marginal return from investing offshore in period 1 is pr + p2( 1 + r): in equilibrium, by going offshore in period 1, in- vestors earn the risk-free rate in both periods. The expected marginal return from investing onshore in period 1, V, , is

+ Pr (c, = R, not binding) ]

= R, binding] )Pr(c2= R, binding)

where I' = 1 if Kl > K;(p) and zero other- wise, and y '' stands for either y or y R, depending on whether c, = F or c, = R.

The term -p2yc1[(r + r)(1 -~[l-1'1) T~PKY- 'I1] in the last line of (4) captures the "political risk" faced when investing in a country subject to potential capital controls. If the probability of controls in period 2, ycl,is zero, then the period-1 marginal product of on- shore capital, PKf- ' p,, equals the offshore rate r, as both onshore and offshore capital yield pr + p2( 1 + r) . In contrast, when the probability of controls in period 2 is positive, then the stock Kl falls in period 1, thus raising the period-1 onshore return flow (the marginal product of capital) above the offshore rate.

The probabilities and are obtained by Bayes' rule from the prior probability that c, = R, conditional on the policy observed in period

1. To clarify the updating process, denote by Rl the set of types imposing controls in period 1, and by R, the set of types imposing controls in period 2. These sets are defined by Rl = Rl(Ko) G {x: $,(KO, X) > 0) and R2 G R2(KI)= {x: I,!J~(K,, X) > O}, where 4, is defined in (3), and $, is similarly defined in

(7) below. Also, denote by G(R1) = SRI dG(x) the prior probability of c, = R, by G(R2) = SRqdG(x) the prior probability of C, = R, and by G(Rl n R,) = SRInR2

dG(x) their joint probability. Then, by Bayes' rule,

Equation (5) illustrates the effects of the persistence of government policies on the per- ceived probability of period-2 controls. The update factor G(Rl n R,)/[G(R,). G(R,)] multiplies the prior G(R2) : should capital- control decisions be independent across peri- ods, then G(Rl n R,) = G(Rl)'G(R2), and the posterior probability of period-2 con- trols, y R, would equal its prior, G (R2). When G(R, n R,) > G(Rl). G(R2), instead, then capital controls provide an unfavorable signal of future policies, as governments imposing controls in period 1 are also more likely to do so in period 2. In this case, the posterior prob- ability of period-2 controls rises above its prior. The intuition behind the upgrade of is symmetrical.

Next, note that V, is a continuous and de- creasing function of K, , for both c, = F and c, = R, going from infinity for Kl -t 0, to p2r for KI -+ w.I6 The unconstrained period- 1 profit-maximizing capital stock, KT = KT (1-1, ,KO, cI ), is then defined as the unique solution for KI of

Finally, since the term multiplying ycl in

(4) is positive (offshore returns exceed on- shore returns with binding controls), then any rise in yC1reduces the return to Kl and, hence, KT itself. This key link underlies our signaling equilibrium: when governments evaluate policy options at the beginning of period 1, they know that actions leading to a higher perceived probability of controls in period 2 will induce a lower desired capital stock in period 1.

We can now close the model by examining the problem faced by a government of type x in period 1. This problem is summarized by the function $, = $,(KO, x), which defines a type x's expected utility from imposing con- trols in this period 1 as a function of the ex- isting capital stock:

-W(7K2 + xl cl = F)]

where expectations are taken over realizations of 1-11 and p2, and the indicator function I" =

l6 To see this, note that PKf-' falls with K, from infin- ity to zero; that yo is increasing in K,, given KO,for both c, = R and c, = F [since G2(Kl,x) rises with K,; see Section 111-A]; and that the term multiplying yo is posi- tive and continuous.

I(c2= R) equals 1 when controls are in place in period 2 and zero otherwise.

The properties of W(*) assure that lim,+-,+I(Ko, x) = m, and lim.r+,$l(Ko, x) = -pE for all KO > 0. Within these extreme values, the behavior of the (contin- uous) function4, determines government policies in period 1, just as 4, does for period

2: types for which 4, > 0 impose controls in period 1, while types for which 4, 5 0 do not. While the relative generality of our model does not allow us to characterize the solution for period 1 as simply as for period 2,17 we can nonetheless characterize gov- ernment behavior in several important re- spects and study its implications for capital flows. Proposition 1 summarizes an impor- tant property: the adoption of free capital mobility in period 1provides a favorable sig- nal, by reducing the posterior probability of controls in period 2.

PROPOSITION 1: In equilibrium, y F( KO, K:(P, KO,cl = F)) < yRw0, K:(P, KO,cl = R)), for all KO> 0 and p > 0: the proba- bility of controls in period 2 is higher condi- tional on capital controls than on free capital mobility in period 1.

The intuition behind Proposition 1 (the proof of which is in the Appendix) is simple. Since capital-control policies are positively correlated across periods, the observation of free mobility in period 1 provides a favorable signal of future policies (i.e., it reduces the posterior probability of controls in period 2). The formal argument behind this correlation is best made by contradiction, by assuming that policies are negatively (or not at all) corre- lated and noting that in this case low-x types would be more inclined to impose controls in period 1 than high-x types. Indeed if the probability of period-2 controls falls upon ob- serving controls in period 1, then imposing controls in period 1 would always raise the end-of-period capital Kl, and the eventual tax base K2: in bad states of nature (where capital

"In particular, it does not seem possible to rule out, in general, multiple intersections between the function $,(KO,x) (as this goes from +m for x -t -m, to -p< for x -* +m)and the horizontal axis.

controls trap capital above its desired level), and in good states (where the assumed favor- able signal provided by capital controls in- creases the desired stock itself). In this case, however, lower-x types would be more inclined to impose controls (just as they do in period 2), since the concavity of their utility function strengthens their taste for a broader tax base. Hence, observing controls in period 1 would increase, rather than decrease (as as- sumed), the likelihood of controls in period 2.

Based on Proposition 1 and on the solution to the investors' problem, Section IV further discusses the first period's equilibrium and its implications for capital flows.

IV. Properties of the Equilibrium

A. Who Imposes Controls?

Our model embodies predictions on what types of governments and circumstances are likely to lead to capital controls. These predic- tions emerge clearly in period 2, due-to the simple solution available in this case: ceteris paribus, governments with fewer outside re- sources, x, or facing greater temptation to im- pose controls (in terms of a larger, potentially captive, capital tax base, Kl ), are more likely to impose controls. Similar predictions emerge in period 1, although the interaction between direct and indirect effects of x and KO in the signaling equilibrium blurs the impact of small changes in these two variables. Nonetheless, the effects of x and KO eventually dominate: free mobility prevails in period 1 as x -t w or KO-t 0, whereas controls prevail as x -t -w or Kn-tw.

These predictions seem realistic. Although the parameter x could capture any of many fac- tors that affect capital-account policies, the in- terpretation we emphasize, that x represents revenues other than from capital taxation which the government may tap in the future, seems empirically appealing. Both casual ev- idence (showing more frequent use of capital controls in developing countries than in indus- trial countries [see Section I]), and evidence from panel studies (see e.g., Alesina et al., 1994), indicate that countries are more likely to impose capital controls when their expected revenues from financial repression are high relative to expected revenues from other sources (here represented by x).I8 Thus, our model's implication that a larger KO should provide stronger temptation to impose controls must be read in a relative sense: developing countries are more likely to use capital con- trols not because their stock of taxable capital is high in absolute terms, but because it is high relative to other revenue sources they can tap.

B. Capital Flows

Several important properties of the model's equilibrium should be noted. First, as long as KO > 0, a separating equilibrium prevails, whereby governments for which QI =$,(KO, x) > 0 impose controls at the beginning of period 1 and all remaining governments allow free capital mobility.l"igure 2 illustrates the equilibrium in this case, by plotting an illus- trative curve $, (ignore the curve $: and the point xR, for the moment).

We have been unable to rule out the possi- bility of nonmonotonic behavior of the curve $, (although we expect the curve to be de- creasing in most plausible cases). In any case, the essence of Provosition 1 is that. even if the ranges of governments choosing c, = R and c, = F may not be connected, in equilibrium there will be enough probability mass attached to c, = R at low values of x to make capital controls in period 1 an unfavorable signal of x and, hence, of future policies. With this caveat, henceforth we shall refer to low-x tvves as be-

,

&

ing more likely to impose controls in period 1.

Next, observe that the profit-maximizing capital stock ~?(p, KO, F) in the high state pl = p conditional on free capital mobility lies strictly above its restricted-mobility counter- part, KT (p, KO, R) (the stocks are clearly both equal to zero in the low state p, = 0). This

See also Roger Gordon and James Levinsohn ( 1989) and Giovannini and de Melo (1 993), who argue that financial repression in developing countries dominates other forms of taxation that are too costly to organize and administer. Value-added or consumption taxes, for in- stance, require sophisticated methods of assessment, bor- der controls, and other measures that may simply be beyond the reach of many poor countries.

l9 When KO= 0, all governments (other than, trivially, the type x = -m) allow free mobility, a poolin 8 equilibrium prevails in period I, ~:(p,,KO,F) = K, (p, , KO, R),and signaling plays no role in the ensuing capital flow. We focus here on the case of KO> 0.

property clearly reflects the separating nature of the equilibrium: the observation of free capital mobility in period 1 triggers a discrete upward revision in investors' beliefs over gov- ernment types, relative to their prior beliefs, and therefore an ex ante increase in expected returns to Kl; when capital controls are im- posed in period 1, the result is symmetrical. The persistence of policies necessary for this result reflects the fact that low-x governments are more likely than high-x governments to impose controls in both periods.

The wedge K?(p, KO, F) -K?(p, KO, R)

affects the response of capital flows to poli- cies. Since ~?(p, KO, F) > K?(~,KO, R) and K?(O, KO, F) = K?(O, KO, R) = 0, for all KO, then EIIK?(pI, KO, F)1 = .irK?(p, KO, F) >

.irKT(p, KO, R) = E,[K?(~,, KO, R)I, for all KO. Hence, there is always a nontrivial range of initial states KO, for which a policy of free capital mobility causes an expected in- flow (i.e., E,[K?(~, , KO, F)] > KO), even though capital controls would lead to a de- sired outflow averaged across states (KO > El [KT (pI ,KO, R)] ) . Intuitively, governments with few outside resources impose controls as their decision is dominated by welfare under the worst scenario (p = 0, in our simple setup), while high-x governments are influ- enced more evenly by the whole distribution of p,. LOW-x governments impose controls be- cause of the large costs associated with pos- sible outflows, notwithstanding the potential benefits of free capital mobility.

The model also predicts the circumstances under which the choice of an open capital ac- count will more likely cause a capital inflow. The likely outcome depends on the strength of the signal provided by that policy. The im- provement in investors' beliefs over types, upon observation of c, = F, is sharper (and hence an inflow is more likely) the lower was the prior likelihood of an open capital account, and the greater is the extent of policy persis- tence across periods. Clearly, for investors to attribute value to this news, they must attach sufficient importance to future policies: equa- tions (4) and (6) indicate that as the discount factor p becomes small, the signaling effect of policies vanishes, investment converges to its one-shot outcome, and a removal of binding controls on outflows always causes an outflow.

C. The Role of Asymmetric Information

To clarifv the role of asvmmetric informa- tion in ou; model, considkr the case where investors are informed of x at the beginning of period 1. Clearly, there is no scope for sig- naling in this case: investors already know whether the host government will impose con- trols in period 2 at each K, . Desired period-1 capital would then be independent of the re- gime Lie., K?(pI, KO, R) = K?(pI, KO, F)1, and a removal of binding controls on outflows would always cause an outflow. Thus, the pos- sibility of an inflow following the removal of binding controls on outflows rests crucially, in our model, on the signaling role of policies.

Asymmetric-information equilibria also ex- hibit "mimiclung" : some governments exploit informational asymmetries to pool with higher-x types and liberalize capital flows in period 1, a policy they would not have adopted with symmetric information. To understand why this is true, consider the highest x (for example, xR [see Figure 21) just indifferent between controls and no controls in the asym- metric-information case (thus, all types higher than xR strictly prefer free capital mobility). With asymmetric information, xR knows that if it imposes controls, investors willfpool its type with types in a range with upper bound xR (and. therefore. reduce their desired in-

\,

vestment), when forming their best guess of the host government's type. Hence, aside from the benefits from trapping the given stock of capital onshore (which are identical with and without asymmetric information), with asym- metric information the type xR also faces a sig- naling cost when imposing controls in period

1. No such cost arises with symmetric infor- mation: investors already know the host govern- ment to be of type xR. Hence, if xRis indifferent between controls and no controls under asym- metric information, xR must strictly prefer con- trols under symmetric information. Combining thls observation with the fact that the symmetric- information case features a simple split between governments imposing controls in period 1 and governments not i~nposing controls [see Figure 2, where +:(KO, x) denotes the gains-from- controls function in the symmetric information case], then the types choosing free mobility with symmetric information form a strict subset of those choosing fiee mobility with asymmetric inf~rmation.~~

Thus, in our model, incomplete information about government attitudes toward capital mobility yields a bias toward liberal markets.

D. Extensions

The model of the previous sections is highly stylized, and several extensions could be con- sidered. The main problem in pursuing some of these may be the loss of tractability: sig- naling models are notoriously hard to solve, and most of the signaling literature has resorted to simplifications such as quadratic loss functions, two-period or two-type models, and the like, in order to obtain tractable solutions. In other cases, augmenting the model to en- dogenize some parameters may not justify the cost of blurring its message. This is likely to be the case, for instance, for more structural models of the motives and costs of capital con- trols to the extent that these yield reduced forms similar to those that we have simply as- sumed (whereby governments have an imper- fectly known taste for domestic capital but suffer a cost from imposing controls).

In some respects, our model is less restrictive than it may appear at first sight. The cost of controls E, for instance, could be al- lowed to vary across periods, to differ across

'O The argument showing that @: is decreasing in x is similar to that for Proposition 1: with symmetric infor- mation, there is no role for signaling, and KT (p,, KO, R) = KT(p,,KO.F). Then, controls can only increase K, and K2;hence, low-x types, which benefit more from a larger tax base, are more inclined to impose controls in all periods.

governments, or to be incurred only when con- trols are binding (e.g., to rise as a function of the gap between onshore and offshore returns). When the cost < is viewed as govern- ment specific, the model's prediction that governments with minimal outside resources are more likely to impose controls on capital outflows must be viewed as a ceteris paribus prediction: governments less able (or willing) to tap outside revenues are more likely to im- pose controls than are governments facing similar costs of imposing controls, but with easier access to outside revenues.

A positive cost <, however, is essential; otherwise the trade-off faced by governments in their policy decision di~appears.~' Therefore, in situations where controls are seen as beneficial (e.g., to insulate domestic markets from external shocks), other costs of disrupt- ing capital mobility ought to be introduced in the model for the government to face a mean- ingful decision problem.

Another simplifying assumption is that the capital stock invested onshore in period 1 remains intact until period 2. When Kt is viewed as physical capital, this assumption is equivalent to that of no capital depre~iation;~~

writing the capital-control constraint as K, 2 K, is also equivalent to assuming controls to be fully ef- fective. Neither of these assumptions is very re- alistic. However, it is easy to extend the model by rewriting the control constraint as K, 2 (1 6) K, -, ,and equations (2) -(4) with ( 1 -6) Kl in place of the stock Kl inherited from period 1. All quahtative results remain unchanged, as long as 6 < 1, that is, as long as capital does not depreciate fully from period to period, and con- trols are at least partially effective.In the degen- erate case of 6 = 1, investors need not be concerned with capital controls: their principal is fully lost (or fully transferrable offshore) in a single period anyway.In the more reahstic case of 0 < 6 < 1, instead, controls would'be effec- tive only in the short run. This is sufficient, how- ever, to make investors afraid that a stock ( l

'I All governments would impose controls in period 2 and, lacking signaling or direct cost of doing otherwise, also in period 1.

''When K, is regarded as a financial claim on returns from physical capital, the same assumption is equivalent to that of no default risk.

6) K, may remain trapped onshore earning a low return, leading to a signaling equilibrium of the type studied above.

The payoffs to a more general model may be significant, but tractability problems seem overwhelming, when extending the analysis to a multiperiod model. It is clear from Subsec- tion 1II.B that our model has a recursive na- ture, and that government and investors would face a very similar problem in each period of a repeated game of duration T. However, whereas we could exploit in our solution for period 1 the semiclosed form solution avail- able for period 2, that strategy would only help for period T-1in a multiperiod model. Nev- ertheless, the two-period case has implications for some results one may anticipate for the multiperiod case.

First, in our two-period model, while a true "liberalization" (in the sense of a re- moval of capital controls following their en- dogenous imposition) can occur in period 2, a period-1 liberalization would be condi- tional on inheriting a regime of restricted capital mobility from period 0. Clearly, a multiperiod extension could produce richer patterns of liberalizations and reimpositions of controls, particularly when combined with a more general treatment of technology. In this respect, allowing for a continuous dis- tribution for the technology shocks would needlessly complicate the model, but allow- ing for serially correlated shocks would have interesting implication^.^' For instance, the stronger the correlation of the shocks, the more information on future tax revenues the government obtains from current shocks. A high correlation and a high value of pwould virtually eliminate the need for controls in response to a positive technology shock: there is no need to insure against bad states of nature if such states can be ruled out a priori. In response to the arrival of bad news on future productivity, the converse is true.

"In our two-period model, this extension would only require specification of the transition probabilities for the technology shocks as conditional on previous states [e.g., ?rp =Pr(p, = pip,-, = p), ?rO = F'r(p, = = 0))

with 1 -?rp and 1 -?rO defined accordingly]; then, ?rp >

> noyields positively correlated shocks. All results would remain qualitatively unchanged.

We expect policies to respond to the same incentives in a multiperiod model: govern- ments with low outside resources, which ex- pect a narrow tax base in the future (because they expect a series of bad productivity shocks) would initially try to trap capital on- shore by imposing controls. Subsequent arrival of good news on future tax revenues, in the form of a large, highly correlated tech- nology shock, would make these govern- ments liberalize capital flows.

Finally, a multiperiod model may make the impact of reputational effects on capital flows even more dramatic. As discussed above, there are governments that would impose restrictions if their type were known, but whch would allow free capital mobility under asymmetric infor- mation in order to be perceived as more likely to adopt liberal policies in the future. The intu- ition from our two-period and other signaling models suggests that these reputational effects could be long-lived.

V. Concluding Remarks

We have presented a model in which gov- ernments can use policies affecting capital mobility to signal a favorable future fiscal sit- uation. In our model, governments with the most to lose from a capital outflow are more likely to fall prey to the temptation of trapping capital onshore; governments with less to gain from a capital inflow are more likely to with- stand such temptation and to allow free capital mobility. Investors recognize these incentives and the persistence of policies affecting capital flows: governments liberalizing capital flows today are more likely not to impose controls tomorrow, and vice versa. Ironically (but in- tuitively), governments with less need for a large tax base are more likely to experience a capital inflow. These predictions are consistent with the observed experience of a number of countries that have liberalized their capital ac- counts. Our model suggests that those policy shifts be viewed as enhancing the credibility of those countries as hosts for foreign invest- ment. The model suggests that a desired inflow upon liberalization is fully consistent with a desired outflow conditional on a repressed capital account.

While capital controls are motivated in our model by their role in broadening the domes- tic tax base, the model's insight should extend to related problems. The main alternative would be to explore the role of capital con- trols to defend an exchange-rate target. We expect this motive to lead to similar predic- tions on the response of investment to capital- control policies. Investors fear being trapped onshore earning a low rate of return. News suggesting lower likelihood of controls in the future would be welcomed with an inflow. Governments with less to lose from a balance- of-payment crisis would try to signal their commitment to free capital mobility by ex- posing themselves to greater chances of a cri- sis. In so doing, they would differentiate themselves from those governments that can- not afford to take chances, thus validating the signaling content of the liberalization.

PROOF OF PROPOSITION 1:

The proof, by contradiction, verifies that, under the converse assumption, 4,(KO, x) falls with x for all KO. Let K:' = ~:(p, KO, F), K? = KT (p, KO, R), and K; = K; (p2). Let x;(K,) be the unique x indifferent between controls and no controls at the beginning of period 2, given K,, and recall that nand 1 nare the probabilities that p, = p and p, =0, respectively, which may depend on { p, -,, p, -*, . . } . We have three cases defined by the relationship between KO and p:

1. Low KO(orhigh p): KO 5 K:' c: KP.

(a) For x 5 x;(K,) (ths is the range of x

that would impose controls in period 2 at KO),

with expectations taken over p2, condi- (a) For x 5 x;(K:), the solution is the tional on p, . same as (Al).

(b)
For x;(K,) 5 x 5 x;(KP), (b) For xT(KP) 5 x -IxT(Ko),
(c)
For x;(K:) Ix CI x;(Kk), (c)For x;(Ko) 5 x 5 x:(Kk), the solu- tion is the same as (A3); and

(A31 JJI(Ko,x) (d) for xT(K7) 5 x, the solution is the same as (A4).

3. High KO(or low p): K: 5 Kk < KO.

(a)
For x < x;(KT), the solution is the same as (Al);
(b)
for xT(K:) 5 x 5 xT(KP), the solu- tion is the same as (A5).
(c)
For x:(Kk) 5 x -= xT(Ko),

(d) For x;(Kk) -.r x,

(A41 $1 (KO,x)

(d) Finally, for x;(Ko) 5 x, the solution is the same as (A4).

+ (1 -n)p2El w(T.K,* + x)

1 Now, by (4) and (5), if Y" 2 yR, then

K: Kk. Then, El[W(* 1 c, = R)] 5

-w(rK,*+x)-- . E,[W(*I c,= F)] for all ranges of x and for Pall cases 1-3. with strict ineauality over

'1

ranges with positive probabili;y. Grther

2. Intermediate KO(orintermediate p): K: I=-more, the difference between the two sides KO5 K?.-of the inequality is decreasing in x, due to

the concavity of -W (*), which implies that sufficiently low x's impose controls in pe- riod l, while higher x's do not. Since this is also true in period 1, then -y < -yR, a contradiction.

REFERENCES

Aizenman, Joshua and Guidotti, Pablo. "Capital Controls, Collection Costs and Domestic Public Debt. " Journal of International Money and Finance, February 1994,13(1), pp. 41-54.

Alesina, Alberto; Grilli, Vittorio and Milesi- Ferretti, Gian Maria. "The Political Econ- omy of Capital Controls," in Leonardo Leiderman and Assaf Razin, eds.,Capital mobility: The impact on consumption, in- vestment, and growth. Cambridge: Cam- bridge University Press, 1994, pp. 289-

321.

Alesina, Alberto and Tabellini, Guido. "External Debt, Capital Flight and Political Risk." Journal of International Economics, November 1989,27(3-4), pp. 199-220.

Bacchetta, Philippe. "Liberalization of Capital Movements and of the Domestic Financial System." Economica, November 1992, 59(236), pp. 465-74.

Backus, David; Kehoe, Patrick and Kydland, Finn. "International Real Business Cycles." Journal of Political Economy, August 1992, 100(4), pp. 745-75.

Banda, Ariel and Santo, Michele. "Una Aprox- imacion Empirica a1 Estudio de 10s Susti- tutos del Dinero en Uruguay." Monetaria, JulyISeptember 1983, VI (3), pp. 301-

14.

Barro, Robert. "Reputation in a Model of Mon- etary Policy with Incomplete Information." Journal of Monetary Economics, January 1986, 17(1), pp. 3-20.

Bartolini, Leonardo and Bodnar, Gordon. "An Analysis of the Process of Capital Liberal- ization in Italy." IMF WPl92127, International Monetary Fund, Washington, DC, 1992.

Bartolini, Leonardo and Drazen, Allan. "When Liberal Policies Reflect External Shocks, What Do We Learn?" National Bureau of Economic Research (Cambridge, MA) Working Paper No. 5727, 1996; Journal of In- ternational Economics, 1997 (forthcoming).

de Melo, Jaime. "Financial Reforms, Stabili- zation, and Growth Under High Capital Mo- bility: Uruguay 1974-83." World Bank DRD Discussion Paper No. 138, Washing- ton, DC, 1985.

Dooley, Michael and Isard, Peter. "Capital Controls, Political Risk, and Deviations from Interest-Rate Parity." Journal of Po- litical Economy, April 1980, 88(2), pp. 370-84.

Fischer, Stanley. "Dynamic Inconsistency, Co- operation and the Benevolent Dissembling Government." Journal of Economic Dy- namics and Control, February 1980, 2(1), pp. 93-107.

Frenkel, Jacob and Razin, Assaf. Fiscal policies and the world economy. Cambridge, MA: MIT Press, 1987.

Fudenberg, Drew and Tirole, Jean. "Perfect Bayesian Equilibrium and Sequential Equi- librium." Journal of Economic Theory, April 1991,53(2), pp. 236-60.

Giavazzi, Francesco and Giovannini, Alberto.

Limiting exchange rate jexibility: The Eu- ropean Monetary System. Cambridge, MA: MIT Press, 1989.

Giavazzi, Francesco and Spaventa, Luigi. "The 'New' EMS." Centre for Economic Policy Research (London) Discussion Paper No. 369, 1990.

Giovannini, Alberto and de Melo, Martha. "Government Revenue from Financial Re- pression." American Economic Review, September 1993, 83(4), pp. 953-63.

Gordon, Roger and Levinsohn, James. "The Linkage Between Domestic Taxes and Bor- der Taxes." Mimeo, University of Michi- gan, 1989.

International Monetary Fund. Exchange arrange- ments and exchange restrictions. Washington, DC: International Monetary Fund, 1995.

Labsn, Raul and Larrain, Felipe. "Can a Lib- eralization of Capital Outflows Increase Net Capital Inflows?" Universidad Catolica de Chile Working Paper No. 155, 1993.

Lucas, Robert. Models of business cycles. Oxford: Blackwell, 1987.

Mathieson, Donald and Rojas-Suarez, Liliana. Liberalization of the current account, Occasional Paper No. 103. Washington, DC: International Monetary Fund, 1993.

Mendoza, Enrique. "Capital Controls and the Gains from Trade in a Business Cycle Model

I54 THE AMERICAN ECONOMIC REVIEW MARCH I997

of a Small Open Economy." IMF Staff Pa- pers, September 1991, 38(3), pp. 480-505.

Obstfeld, Maurice. "Capital Flows, the Current Account, and the Real Exchange Rate: Consequences of Liberalization and Stabi- lization." National Bureau of Economic Research (Cambridge, MA) Working Paper No. 1526, 1984.

OECD. Economic surveys: New Zealand. Paris: Organization for Economic Cooper- ation and Development, 1987.

PCrez-Campanero, Juan and Leone, Alfredo. "Liberalization and Financial Crisis in Uruguay, 1974-87," in V. Sundararajan and T. Baliiio, eds., Banking crises: Cases and is- sues. Washington, DC: International Mon- etary Fund, 1991, pp. 276-375.

Reserve Bank of New Zealand. Bulletin. Wellington: Reserve Bank of New Zealand, Jan- uary 1986a.

.Bulletin. Wellington: Reserve Bank of New Zealand, July 1986b.

Rogoff, Kenneth. "Reputational Constraints on Monetary Policy." Camegie-Rochester Con- ference Series on Public Policy, Spring 1987,26, pp. 141-81.

Schadler, Susan; Carkovic, Maria; Bennett, Adam and Kahn, Robert. Recent experiences with surges in capital inJows ,Occasional Pa- per No. 108. Washington, DC: International Monetary Fund, Washington, DC, 1993.

Vifials, JosC."Spain's Capital Account Shock." Centre for Economic Policy Re- search (London) Discussion Paper No. 477, 1990.

Williamson, John. ''On Liberalizing the Capital Account," in Richard 07Brien, ed., Finance and the international economy 5: The AMEX Bank Review Prize Essays.Oxford: Oxford University Press, 1991, pp. 136-46.

Comments
  • Recommend Us