Currency Unions
📄 Summarized from the full manuscript · Human-reviewed for faithfulness before publication
In brief
Which countries gain the most from giving up their own currency? Alesina and Barro combine a trade model, where sharing a currency lowers trading costs, with a Barro-Gordon inflation-commitment model, where a country that cannot credibly commit to low inflation benefits from anchoring to one that can. The answer: small, trade-intensive economies whose business cycles move with a large, low-inflation anchor gain the most -- and as the number of sovereign states rises, the model predicts the number of currencies in circulation should rise less than proportionally, or even fall.
What this paper finds — and why it matters
This 2002 Quarterly Journal of Economics paper by Alberto Alesina and Robert Barro builds a formal model of which countries should give up their own currency and adopt another country’s, combining a real trade/country-size model with a Barro-Gordon (1983) rules-versus-discretion monetary model. In the real side, competitive final-goods producers combine differentiated, monopolistically-supplied intermediate goods subject to an iceberg trading cost between countries; a reduction in this cost, including via a shared currency, raises consumption in both trading partners, and does so more for a country that trades a lot with its partner and for a country that is small relative to the world number of varieties. On the monetary side, a country that cannot itself commit to low inflation faces a discretionary inflation bias (the Barro-Gordon outcome); linking to an anchor country that can commit avoids this bias but costs the client both the ability to react to its own idiosyncratic output shocks and exposure to relative-price shocks between the two countries’ consumption baskets. Combining the two channels, the paper shows that a country gains more from adopting an anchor’s currency the harder it would find it to commit to low inflation on its own, the more it trades with and the more its business cycle is correlated with a large, low-inflation anchor, and the smaller it is. The paper also shows that a smaller anchor is less able to sustain a monetary-policy commitment against ex post pressure from a large client (the “Russia-Estonia” problem), and works out how an anchor’s optimal policy changes if clients compensate it for partially accommodating their disturbances. Finally, aggregating these individual decisions across many countries arranged by distance/trade intensity, the model predicts that currency unions form among adjacent, closely-linked countries, and that as the number of independent countries in the world rises, the equilibrium number of currencies in circulation can rise less than proportionally, or even fall.
Summary of a classic paper, AI-assisted and human-reviewed. See the linked original for the authoritative claims and full conditions.
Questions & answers
Q1. How does the real side of the model link country size, trading costs, and the gains from a common currency?
Final-goods producers in each of two countries combine differentiated intermediate inputs via a Dixit-Stiglitz/Ethier-type production function (eq. 1); shipping an intermediate good across the border destroys a fraction b of it (an iceberg cost), while the final consumer good trades costlessly, so its price is equalized and normalized to one (Sec. II.A, pp. 412-413). Consumption per person in country I is proportional to [N^I + (1-b)^(a/(1-a)) times N^II] for equal markups (eq. 8, p. 416). The paper derives that a lower trading cost b raises consumption in both countries, that the gain is larger the more a given pair of countries already trades (provided the underlying goods are relatively close substitutes, a greater than 1/2), and that a smaller country (lower N^I for given world N) gains more from reducing b (Sec. II.B, pp. 416-418) – the three comparative statics that later determine which country-pairs want to share a currency.
Q2. How is the monetary side of the model set up, and what is the discretionary (autonomous) outcome?
The paper writes a country i’s policy loss as a quadratic in actual inflation plus a term resembling an expectational Phillips curve, combining actual inflation, squared actual inflation, and a squared gap between (a scaled measure of) unexpected inflation and a disturbance z_i plus eta_i, where z_i is increasing in the average markup distortion and eta_i is a mean-zero shock to that markup (Sec. II.D, eq. 16, pp. 422-423) – a reduced form the authors derive heuristically from the real model’s markup-pricing distortion. Under discretion, solving period-by-period as in Barro and Gordon (1983) yields a positive inflation bias plus a term reacting to the current markup shock, with the bias term eliminated only if the country can commit (Sec. II.D, eqs. 17-18, p. 423).
Q3. What is the expected cost comparison that determines whether a country should “dollarize,” and what do its three terms mean?
Equation (22) (Sec. II.F, pp. 425-426) gives the difference in expected policy cost between staying autonomous and linking to a committed anchor j as a sum of three terms: (i) a term proportional to the square of z_i – the inflation-bias cost avoided by anchoring, which is positive only for countries that cannot themselves commit; (ii) a term in the variance of relative-price shocks between the two countries’ consumption baskets – a cost of anchoring, since the client inherits the anchor’s inflation only up to this noise; and (iii) a term in the variance of the two countries’ markup/output disturbances relative to each other – a cost of anchoring, reflecting the independent-monetary-policy benefit the client gives up. A positive overall value favors dollarization; thus a non-committing country that trades closely with, and has a correlated business cycle with, a low-inflation anchor gains the most from linking to it.
Q4. Under what condition does an anchor committed to a simple fixed rule dominate one using full discretionary feedback, from the client’s perspective?
The paper shows an anchor that commits to a simple noncontingent rule (constant inflation, no feedback from its own shock) is more attractive to the client than a fully optimal, feedback-responsive anchor whenever the correlation between the countries’ shocks is less than one-half (Sec. III.A, pp. 426-427) – because when shocks are weakly correlated, the anchor’s optimal feedback response to its own disturbance mostly imports noise rather than useful co-movement into the client’s inflation.
Q5. How does the model change when the anchor partially accommodates the client’s shocks in exchange for compensation?
If the anchor instead minimizes a weighted joint loss across itself and the client (weights equal to each country’s share of combined consumption) and the client compensates it accordingly, the anchor’s optimal rule becomes one that partly offsets the client’s disturbance in proportion to the client’s size relative to itself (eqs. 24-25, Sec. III.B, pp. 427-428). The compensation the client must pay the anchor for this accommodation (eq. 26) is smaller the smaller the client is relative to the anchor, which makes dollarization with a large anchor more attractive to a small client even after paying for accommodation (Sec. III.B, pp. 428-429) – though the paper notes a large client relative to a small anchor (its “Russia-Estonia” example) may simply be unable to make the anchor’s commitment credible against ex post pressure, regardless of compensation.
Q6. What configurations of currency unions does the model predict in equilibrium across many countries?
Defining an equilibrium as a configuration from which no country wants to leave its current union (or lack thereof) to join another, the paper proves that if countries i and i+2 belong to the same currency union, so must country i+1 – currency unions form among geographically/economically adjacent countries (Sec. IV.B, Configuration of Equilibria, pp. 431-432). Depending on parameters, equilibria range from every country keeping its own currency, to two large currency unions splitting the world, to several multi-country unions coexisting with some fully independent countries – and, if trade gains alone (absent any commitment motive) are large enough, even countries that could otherwise commit on their own may pool into their own union.
Q7. What does the model imply about how the number of currencies should change as the number of countries in the world rises?
The paper shows with a worked three-country numerical example that the equilibrium number of currencies need not rise proportionally with the number of independent countries, and can even fall, as trading costs and the benefits of monetary commitment interact with country size (Sec. IV.C, pp. 432-433). The authors connect this to the post-WWII proliferation of sovereign states (from 76 countries in 1947 to 193 by the time of writing) alongside increasing international integration, arguing the “one country, one money” norm should be expected to erode as small, highly integrated states multiply – a pattern the paper suggests was historically masked by the symbolic/sovereignty value attached to a national currency, which recent dollarization and currency-union episodes suggest may be weakening.
Q8. What empirical work by others does the paper draw on for the magnitude of the trade benefits of currency union?
The paper cites Rose [2000], Frankel and Rose [2002], and Glick and Rose [2001] as finding that the trade-boosting effect of sharing a currency “may be substantial, well beyond the effects from fixed exchange rates,” while noting that Tenreyro [2002] finds smaller effects (Introduction, p. 411) – the paper’s own contribution is the theoretical linkage between this trade channel and the separate monetary-commitment channel, not a new trade-effect estimate of its own.
Key terms in this paper
Definitions below follow the paper's own usage.
- Dollarization/anchor-choice criterion
- The paper's three-term expected-cost comparison (Sec. II.F, eq. 22) for whether country i should adopt committed country j's currency: a benefit from avoiding i's own discretionary inflation bias (relevant only if i cannot itself commit), a cost from the variance of relative-price shocks between the two countries' consumption baskets, and a cost from the variance of the two countries' idiosyncratic output/markup disturbances relative to each other (the independent-monetary-policy benefit forgone).
- Seignorage-sharing (compensated accommodation)
- The paper's extension (Sec. III.B, eqs. 24-26) in which an anchor country partially adjusts its inflation rate to offset a client's disturbances, in proportion to the client's share of their combined consumption, in exchange for compensation from the client; the compensation required is smaller the smaller the client is relative to the anchor, but a client that is large relative to the anchor may be unable to make the anchor's commitment credible at all.
- Adjacency result for currency-union formation
- The paper's equilibrium configuration theorem (Sec. IV.B): if countries i and i+2 belong to the same currency union, country i+1 must also belong to it -- currency unions form among countries that are adjacent in the sense of being close in distance, trading heavily, and having correlated business-cycle shocks, not among arbitrarily dispersed sets of countries.