2 1695 DEVELOPMENT BRIEF Number 54 The World Bank May 1995 Sunk costs: Why exporters tured goods during the 1980s. In- Sun k costs: Wvhy exporters cluded in the sample are 650 plants that were operating in each of the remain exporters nine years. How to get companies to export has long been a The hard part: Getting in puzzle. What works today will not necessarily work The model shows that once a pro- tomorrow. "Sunk costs" may have something to do ducer has met the costs of entering an export market, it will remain in with it the market as long as its operating D epending on the time and Although the sunk cost hypoth- costs are covered. If that's the case, the place, economists have esis seems reasonable, it has not changes in government policy, ex- found different, sometimes been easy to validate. Shifts in trade change rates, or prices in foreign conflicting, answers to the question flows can be correlated with markets can permanently alter trad- of what works when a country changes in export policy. But do the ing patterns and national exports. wants to stimulate exports. As a re- figures reflect increased or de- For example, policy changes that sult, policymakers have gotten little creased sales by the same exporters entice companies into exporting can guidance on how their national or an increase or decrease in the continue to increase the flow of ex- manufacturers might respond to number of companies exporting? A ports even if the policy is later one incentive or another. A trade or recent World Bank research project changed. But if businesspeople sus- exchange rate policy that works has developed an econometric pect that the policy will be changed marvelously in one country often in the future, they will shy away. fails in another-and sometimes Companies will not start exporting even in the same country when if they conclude that export incen- tried a few years later. If you can start tives might disappear after they Some researchers have concluded firms exportin have paid the entry fee. Stable poli- that the effect of changes in markets rm P expotingcies can bring more companies into or in export policy can be masked they are likely to exporting than can transitory or offset by the potential exporters' changes in exchange rates. entry or "sunk" costs-the largely continue exporting A plant that exported last year is one-time expense of gathering in- likely to export again this year. But formation on foreign markets, up- after two years out of the export grading product quality, changing model that tests theory against the market, the plant is not much more packaging, and establishing mar- real comings and goings of a likely to export than one that has keting channels. If such expenses nation's manufacturers.* The model never done so. Export experience do weigh heavily in export deci- not only reflects the influence of can be lost. Market information sions, they might produce "hyster- sunk costs, but is sensitive to other goes out of date rapidly-a new esis" in international trade. Put in conditions, including a plant's prior exporting campaign means a new, English, if you can start firms ex- exporting experience. costly effort to accumulate porting-say, by setting favorable The model is based on the export information. exchange rates-they are likely to patterns of plants in four major Sunk costs are not the only factors continue exporting even when the exporting industries in Colombia in export decisions. Profit expecta- rates are no longer favorable. over a nine-year period, 1981-89. tions also are important. Rising and The industries-food, textiles, falling entry costs will, of course, 'For more details, see Mark J. Roberts and James R. Tybout, paper products, and chemicals- influence the calculation of net "An Empirical Model of Sunk Costs and the Decision to accounted for nearly 60% of profits. But so will other factors, in- Export," Policy Research Working Paper 1436, World Bank, Washington, DC, 1995. Colombia's exports of manufac- cluding the strength of the cur- rency. Many Colombian companies apparently entered the export mar- Declining peso = climbing exports ket in 1986 and 1987 because they A Colombian export subsidy meant to difficult to import the raw materials anticipated several years of peso expand trade did not help. A depreci- or capital goods that manufacturers depreciation. (They were right.) ating peso obviously did. The figure, needed to compete intemationally. Profit calculations also appear to be which charts export activity against The value of exports and the number the reason that plants that are large, the subsidy and the exchange rate, is of companies exporting declined until old, abased on census data for 19 Colom- 1984. By then, the peso was weaken- old, and owned by corporations are bian industries accounting for 96% of ing, in part because of central bank much more likely than smaller the country's exports. Hlegal exports, intervention. As it continued to de- firms to be exporters. The large an inflow of foreign capital, and a cof- cline, exports rose sharply even plants probably have cost advan- fee boom led to a steady appreciation though export financing was difficult tages, or are simply further along of the peso from the 1970s into the and businesspeople had to be con- the export learning curve. Larger early 1980s. Exporters were further vinced that lobbyists for protection- the export learning curve. Larger hurt by trade policies that protected ism would not succeed in reversing a plants also are more likely to pro- domestic industry, but also made it move toward trade liberalization. duce the type of product that can be sold in volume to a few foreign Colombian exports of manufactures responded largely to exchange customers-lowering ongoing mar- rate movements keting expenses if not entry fees. (Index: 1981 = 100) Not surprisingly, location also 300 counts. Exporting companies are more likely to be in the port cities, 250 Real value of exports Baranquilla and Cartagena, than in / Andes-locked Bogota. 200 / Lessons for policymakers 150 _ ' - Exporting plants The combined effects of sunk costs, type of plant, and even type of --- - Exi ownership mean that industry's re- _ _ sponse to changes in policy or macroeconomic factors will vary 0 from time to time and from place to 1981 1982 1983 1984 1985 1986 1987 1988 1989 place. The effect of a change in Source: Mark J. Roberts and James R. Tybout, 'An Empirical Model of Sunk Costs and the Decision policy will depend on the number to Export," Policy Research Working Paper 1436, World Bank, Washington, DC, 1995. of plants already exporting, on whether managers perceive the With all these forces at work, port sales, governments can profit- changes as temporary, and on the policymakers still lack a clear direc- ably abandon export subsidies al- cost of entry. In turn, the cost of tive on how to stimulate interna- ways perceived as temporary and getting into a market will vary with tional trade. They should recognize, concentrate instead on providing the type of product, the amount of however, that they need two pro- market intelligence, improving ex- information that exporters have grams: one to encourage existing port infrastructure, and replacing about the foreign market, and the exporters and another to entice uncertainty with policy stability conditions in the importing coun- newcomers into international trade. and consistency. try. An expanding market is easier And if getting into exporting is a and cheaper to enter. greater problem than expanding ex- Developnleel Brefs are issued by the World Bank to inform the media, business, academic, and governmentpolicv communities aboutdevelopmentpolicy analyses and results from the Bank's research activities. They are draw-n fromthe workof individualBankresearchersand do not-necessarily representtheviewsof theWorld Bank and its membercountries-and should notthereforebe attributed to the World Bank or its affiliates. Briefs are issued periodically by the Research Advisory Staff, Development Economics Vice Presidency, The World Bank, 1818 H Street, NW, Washington, DC 20433. Tel: (2()2)473-3984, Fax: (202)477-0955. Briefs are not copyrighted and may be reproduced with the appropriate attribution.