does economic development lead to mangrove loss? a cross-country analysis

DOES ECONOMIC DEVELOPMENT LEAD TO MANGROVE LOSS?A CROSS-COUNTRY ANALYSIS

EDWARD B. BARBIER and MARK COX*

Mangroves line one quarter of the world's tropical coastlines, and approximately117 countries and territories have mangrove resources within their borders. Althoughover recent years mangrove deforestation has occurred at a phenomenal rate worldwide,there have been few economic studies of the underlying causes. The article attempts suchan analysis and particularly examines the role of economic development, with specificreference to those activities that may result in mangrove deforestation, in determiningthe area of mangrove left within a country. The article develops a model of economicactivity and mangrove conversion. From this model, a relationship is establishedbetween remaining mangrove area, economic activity, and other important causativefactors. The mangrove area relationship is estimated empirically for a cross-section of89 countries. Results show that shrimp aquaculture and agriculture are significantlyassociated with mangrove loss across all countries, whereas the higher the level of GDPper capita the more mangrove area remains. The number of protected areas, length ofcoastline and political stability were also important in determining the remainingmangrove area of a country. (JEL O13, Q22, Q23, Q24)

I. INTRODUCTION

Mangrove, or mangal, systems are the sub-tropical and tropical equivalents of the tempe-rate coastal and estuarine salt marsh system.

They are essentially forest-based systems thattolerate salt and occupy the intertidal zonebetween land and sea. Although mangrovesare generally found within 25� north and southof the Equator, they can be found in somenorthern latitudes as high as 32� (Maltby,1986).

Mangroves line one quarter of the world'stropical and subtropical coastlines, covering anarea of between 190,000 and 240,000 km2 glob-ally (Kelleher et al., 1995). Approximately 117countries and territories have mangroveresources within their borders (WCMC, 1994).Indonesia has the largest area of mangroveforest, estimated at 4.5 million ha. Nigeria,Australia, Mexico, and Malaysia have thenext largest areas of mangrove forest estimatedat around 1 to 2 million ha (WRI, 1996). Man-groves are very important to many tropical andsubtropical countries, because they serve toprotect coastlines from tidal waves, sea ero-sion, and hurricanes. Furthermore, they arehighly productive natural ecosystems and

ABBREVIATIONS

FAO: Food and Agriculture Organization of the

United Nations

GDP: Gross Domestic Product

*Work on this paper was undertaken as part of theproject Demographic and Economic Factors DeterminingCoastal Land Conversion into Commercial Shrimp Farms,Thailand, funded by the Population, Consumption andEnvironment Initiative of the John D. and Catherine T.MacArthur Foundation. The authors would like to thankthe MacArthur Foundation and the PCE director,A. Kumar, for support. The authors also thank collabor-ating researchers on the project at the Institute for SocialandEconomicPolicy,Bangkok,inparticularS.Sathirathai,R. Tokrisna, S. Aksornkoae, W. Sungunnasil, I. Sarntisart,and S. Suwannodom, who all provided valuable help andadvice on all aspects of this research. Finally, the researchassistance of S. Chanyaswad is acknowledged. A version ofthis article was presented at the Ecosystem, Conservationand the Environment Session, Western Economics Asso-ciation International Meetings, San Francisco, July 7,2001. The authors thank Paula Despins, Darwin Hall,Denise Stanley, and three anonymous referees for invalu-able comments.

Barbier: John S. Bugas Professor of Economics,Department of Economics and Finance, Universityof Wyoming, Laramie, WY 82071-3985. Phone 1-307-766-2358, Fax 1-307-766-5090, E-mail [email protected]

Cox: Research Associate, Centre for Environmental andDevelopment Economics, Environment Department,University of York, Heslington, York YO10 5DD,UK. Phone � 0113 2709498, E-mail [email protected]

418

Contemporary Economic Policy(ISSN 1074-3529) DOI: 10.1093/cep/byg022Vol. 21, No. 4, October 2003, 418±432 # Western Economic Association International

provide nutrients and shelter to many commer-cially importantaquaticorganisms(MitschandGosselink, 1993; Mooney et al., 1995; WCMC,1994; WRI, 1996).

Today, mangroves are one of the world'smost threatened ecosystems and are rapidlydisappearing in many tropical countrieswhere they were once abundant. For example,Malaysia may have lost 17% of its mangrovearea between 1965 and 1985, India as muchas 50% between 1963 and 1977, and thePhilippines as much as 70% between 1920 and1990 (WRI, 1996). Many of the other countriesin Asia, Latin America, and Africa have lostbetween 30% and 70% of their mangrove areain the last 30 to 40 years (Spalding et al., 1997;WRI, 1996). In some countries, such asThailand, the rate of mangrove loss has beenmore recent yet extremely rapid. Over 1975±93 the area of mangroves in Thailand has vir-tually halved, from 312,700 ha to 168,683 ha(Sathirathai and Barbier, 2001).

Although increasing population pressurein coastal areas and overharvesting of timberand other wood products are contributing tothe destruction of mangrove forests, in recentyears a more significant cause appears to be thedemand for land by key primary sector eco-nomic development activities, such as mining,conversion to salt ponds, and agricultural andaquaculture expansion. By far the most impor-tant of these activities is believed to be theexpansion of aquaculture ponds, especiallyfor shrimp production, into mangrove forests(Aksornkoae et al., 1986; Primavera, 1997;Spalding et al., 1997; WRI, 1996). In recentdecades, shrimpandfishaquaculture is thoughtto have accounted for conversion of 20% to50% of mangroves worldwide (Primavera,1997).1 The growing importance of shrimpfarming to the export earnings of tropical coun-tries may have further exacerbated this pro-blem. Since 1989, shrimp aquaculture hasincreased by over 400%, and its share ofworld shrimp production increased from 5%in 1982 to over 30% more recently (Andersonand Fong, 1997).2

The main concern over mangrove lossworldwide is that it results in severe disruptionto the important ecological and economic func-tions normally performed by undisturbed man-grove systems. In many countries and regions,mangrove deforestation is contributing to fish-eries decline, degradation of clean water sup-plies, salinization of coastal soils, erosion, andland subsidence, as well as release of carbondioxide into the atmosphere (Barbier andStrand, 1998; Naylor et al., 2000; Ruitenbeek,1994; Spalding et al., 1997; WRI, 1996). A posi-tive correlation between mangrove area andoffshore shrimp and fish catches has beendocumented for the Philippines, Malaysia,Indonesia, and Australia (Primavera, 1997).In Thailand, the welfare losses associatedwith the impacts of mangrove deforestationon coastal fisheries in Surat Thani Provincewere estimated to be around US$21±52 perha (Sathirathai and Barbier, 2001).

Although the loss of mangroves and theresulting environmental effects are well pub-licized, there have been few studies of theeconomic causes of mangrove deforestation.For example, a recent article conducts a panelanalysis of the economic and demographicfactors determining the conversion of man-groves in the coastal provinces of Thailandto shrimp farming (Barbier and Cox, 2002).A related article estimates the economiclosses to off-shore shellfish and demersal fish-eries in Thailand resulting from mangroveconversion by shrimp aquaculture (Barbier,2003).

This article represents the first attempt todevelop a cross-country empirical analysis ofthe extent to which economic developmentinfluences mangrove loss worldwide. To dothis, the authors develop a basic model of eco-nomic activity and mangrove conversion, inwhich the demand for land by converting activ-ities leads to mangrove loss. From this model, arelationship is established between remainingmangrove area, the level of economic activityand other important causative factors. Man-grove area is expected to be a decreasing func-tion of aggregate economic activities, such asshrimp aquaculture and agricultural expan-sion, that depend on the conversion or deple-tion of mangroves and to be increasing with

1. Mangrove swamps are considered very suitable forshrimp farming because the areas are flooded with brack-ish, stagnant water that is ideal for aquaculture (Kongkeo,1997).

2. Shrimp aquaculture exports are particularly impor-tant for leading producers, mainly in Asia. For example, inBangladesh shrimp farming contributes 8% to total exportearnings (Raha and Alam, 1997). In Thailand, the total

value of export earnings for shrimp in the late 1990s wasaround US$1±2 billion annually (Jitsanguan et al., 1999;Tokrisna, 1998).

BARBIER & COX: MANGROVE LOSS 419

the amount of environmental protection.Ecological and coastal conditions, the accessi-bility of mangrove areas, and institutionalfactors (such as political stability) shouldalso have an influence. Although the authorswould ideally like to conduct a pooled cross-sectional and time-series analysis of mangrovedeforestation across all countries, to date thebest available data source on mangrove areasworldwide contain only a single-year estimateof mangrove area by country (Spalding et al.,1997). The authors therefore estimate man-grove area relationship empirically through across-sectional analysis of 89 countries. Thearticle also analyzes two important subsam-ples, middle and low-income economies andcountries with mangrove areas greater than25 km2.

The structure of the article is as follows. Thenext section develops an economic model ofmangrove conversion that is the basis for deriv-ing a relationship between remaining man-grove area, the level of economic activity, andother important causative factors. Section IIIdescribes the approach to estimating thismangrove relationship and presents the cross-country analysis of the factors determiningmangrove area worldwide. Section IV dis-cusses and compares the estimation results inmore detail, and section V provides an overallconclusion.

II. A MODEL OF ECONOMIC ACTIVITY ANDMANGROVE CONVERSION

Suppose that there are J different sectors, oractivities, in the coastal area of an economythat depend in some way on the conversionor depletion of mangroves. For example,agriculture, aquaculture, salt production, andresidential or industrial construction wouldinvolve conversion of mangrove lands. Log-ging, charcoal making, tannin production,and other activities that depend on depletingmangrove forest resources could also lead to aloss of mangrove areas. Although these activ-ities may be highly diverse, their productionrelationships could be represented in a similarway, namely, as a function of converted man-grove area and other inputs.

Thus, assume that the aggregate productionof all J economic sectors that lead to mangroveloss can be represented by a single productionrelationship and total output, y. At any time t,

the total stock of 1, . . . K inputs (e.g., labor andcapital) available for producing this output canbe represented by the vector, x. However, onlyx1 of these inputs are used directly in the pro-duction of y. The remaining x2 units are used toconvert mangrove area, A, which of course arealso an input for producing y. Thus the basiccost-minimizing level of production for y canbe defined as

C�w, y� � minx

wx�1�

subject to

y � F�x1, A�,�2�

A � A�x2, a�,�3�

x � x1 � x2:�4�

Equation (2) is the aggregate productionfunction for y, which is assumed to exhibitthe standard properties with respect to its argu-ments, x1 and A. Equation (3) is the relation-ship for the level of mangrove conversion, A,which is an increasing function of the amountof inputs allocated to conversion, x2, and arange of exogenous factors, a, that may influ-ence the accessibility of mangrove areas avail-able for conversion, including roads, coastalinfrastructure, and concentration of popula-tions in coastal areas. Equation (4) indicatesthat all inputs, x, can be used either directlyin production of y or for mangrove conversion,with w being the corresponding vector of inputprices.

It follows from these relationships that theconditional factor demand for any input, xik,can be defined as

xik � xik�w, y, a�, qxik=qwik50,�5�qxik=qwl40,xik=qy40,

i� 1, 2 k� 1, . . .K k 6� 1:

That is each kth input used either directly forproduction, x1k, or for mangrove conversion,x2k, is decreasing in its own price, wk, butincreasing with respect to other input prices,wl, and output, y. The corresponding cost-minimizing vector of inputs that is used for

420 CONTEMPORARY ECONOMIC POLICY

mangrove conversion is defined as x2� x2(w,y, a), and thus (3) can be rewritten as

A�A�w, y, a�, qA=qy40, qA=qa40:�6�Let the total amount of mangrove area that

is remaining in any time period, t, be defined asM. Thus, the total stock of mangrove area attime t available in a country, either for conver-sion or preservation, can be denoted as N>A,or N�A�M. Because this is a static model(for cross-sectional analysis) the authors willassume that N is a given stock and will beaffected by prevailing ecological and coastalconditions, b, of the country, such as lengthof coastline, rainfall, tidal conditions, tempera-ture, water quality, and so forth. In addition,the remaining area of mangrove, M, may beinfluenced by the degree of conservation andprotection effort, g , allocated to their preserva-tion. It follows that

A�w, y, a��M�g��N�b�, qM=qg40:�7�

Rearranging (7) yields

M �M�w, y , a, b, g�, qM=qy50,�8�qM=qa50, qM=qg40:

Equation (8) indicates that the remainingmangrove area in a country should decreasewith the aggregate output, y, of economic activ-ities that depend on mangrove conversion,as well as with factors that increase theaccessibility of mangrove areas, a, but increasewith the amount of environmental protection,g . Ecological and coastal conditions, b, andinput prices, w, should also affect the extentof mangrove areas remaining in a country.

Finally, in recent years, a variety of empiri-cal analyses at both the country and cross-country level have explored the impact ondeforestation of institutional factors, such asland use conflict, security of ownership or pro-perty rights, political stability, and the ` ruleof law'' (e.g., Alston et al., 2000; Barbier andBurgess, 2001; Deacon, 1994, 1999). The mainhypothesis tested is that such institutionalfactors may have important independentinfluences on deforestation, separate fromthe effects of other explanatory economic vari-ables. It seems reasonable that the samehypothesis may hold for the current analysis,

namely, that prevailing institutional condi-tions may be an additional and independentinfluence on the amount of mangrove arearemaining in a country. Denoting z as a vectorof institutional indices, such as measures ofpolitical stability, ownership security, landuse conflict, and the rule of law, then a modifiedversion of (8) is

M �M�w, y, a, b, g ; z�, qM=qy50,�9�qM=qa5 0, qM=qg 4 0:

Equation (9) is therefore the mangrove arearelationship to be estimated through the fol-lowing cross-country analysis

III. DATA AND ESTIMATION APPROACH

To date, the most reliable source of interna-tional mangrove data is the World MangroveAtlas (Spalding et al., 1997). This databasecontains estimates of mangrove area for 89countries, based on various satellite imageryand map sources.3 However, the Atlas reportsonly a single-year estimate of mangrove areafor each country. In addition, because differentsources are used to provide this estimate, theyear in which mangrove area is estimated variesgreatly from country to country. Finally, thecountries included in the database vary consid-erably in terms of the size of their mangrovearea and stage of economic development. Forexample, 22 countries in the sample containmangrove forests of 2,000 ha or less, whereas28 countries have mangrove areas of 1.5 millionha or more. Although the vast majority ofthe countries are low- and middle-incomeeconomies, 14 have gross domestic product(GDP) per capita ranging from $7,000 toaround $23,400.

Table 1 lists the 89 countries from theMangrove World Atlas that were used inthe following cross-country estimation ofequation (9). The year of estimation of man-grove area for each country is indicated in

3. Although the list of countries covered by the WorldMangrove Atlas is fairly comprehensive, there are a coupleof notable omissions, such as Nigeria, which has one of thelargest areas of mangroves in the world (WRI, 1996), andthe United States. Curiously, the Atlas does not reportmangrove area for the entire United States but only forFlorida.


parentheses. In addition, the table distin-guishes between countries with large(> 25 km2) as opposed to small (<25 km2)mangrove areas, and those with high(> $7,000) per capita income as opposed tomiddle or low-income (<$7,000).

To estimate equation (9), the authorsrequired cross-country data for the key vari-ables in the regression that also matchedas best as possible the year of estimation ofmangrove area for each country, as reportedin Table 1. The authors were able to finddata, or close proxies, for output of mangrove-dependent economic activity (y), accessibilityof mangrove areas (a), ecological and coastalconditions (b), environmental protection (g),

and institutional factors (z). Unfortunately, theauthors were unable to obtain suitable data torepresent the input price variable, w.4 Forthe remaining variables in (9) the article usesthe following data sources. Unless indicated

TABLE 1

Countries from World Mangrove Atlas Sample

Low income (<US$7,000 per capita per year)

Low mangrove (<25 km2)

Aruba (80) Grenada (80) Micronesia (69) St. Vincent (80)

Benin (89) Guam (76) Samoa (95) Tonga (72)

Djibouti (85) Mauritania (95) St. Kitts (80) Vanuatu (72)

Dominica (80) Mayotte (85) St. Lucia (84)

High mangrove (>25 km2)

Angola (92) Ecuador (91) Indonesia (89) Senegal (85)

Antigua (91) Egypt (92) Iran (70) Seychelles (78)

Bangladesh (77) El Salvador (81) Kenya (95) Solomon Islands (95)

Belize (92) Eq. Guinea (60) Madagascar (79) Somalia (92)

Brazil (91) Fiji (85) Malaysia (86) South Africa (92)

Brunei Dar (92) French Guiana (79) Mexico (92) Sri Lanka (92)

Cambodia (88) Gabon (94) Mozambique (80) Sudan (92)

Cameroon (85) Gambia (85) Myanmar (95) Surinam (78)

Colombia (85) Guadeloupe (80) Nicaragua (91) Tanzania (89)

Comoros (78) Guatemala (92) Pakistan (93) Thailand (87)

Congo (92) Guinea (80) Panama (88) Togo (95)

Costa Rica (88) Guinea-Bissau (90) Papua New Guinea (70) Trinidad Tobago (80)

Cote d'Ivoire (85) Guyana (96) Peru (91) Venezuela (82)

Cuba (89) Haiti (82) Philippines (87) Vietnam (87)

Dominican Republic (84) India (86) Puerto Rico (78) Yemen (87)

High income (>US$7,000 per capita per year)

Low mangrove (<25 km2)

Bahrain (95) Bermuda (75) Qatar (92) Singapore (90)

Barbados (75) Hong Kong (89)

High mangrove (>25 km2)

Australia (95) Cayman Islands (78) New Zealand (82) Saudi Arabia (85)

Bahamas (92) Japan (86) Oman (88) United Arab Emirates (82)

Note: For each country, the year in which mangrove area is estimated is indicated in parentheses.

Source: Spalding et al. (1997).

4. The preference was to use rural wage rates acrosscountriesas the main inputprice variable,because the avail-able evidence suggests that in many countries rural labor isa major input in both the clearing of mangroves and inactivities such as aquaculture and agriculture that dependon mangrove conversion (Spalding et al., 1997; WCMC1994; WRI, 1996). However, it was not possible to obtaina cross-country data set on rural wage rates for the year ofestimation of mangrove area for each country, as reportedin Table 1. Similar problems occurred for other input pricedata, such as for fertilizers, insecticides, and other key agri-cultural inputs.


otherwise, the data are for the same year ofestimated mangrove area for each countryreported in Table 1.

A. Mangrove-Dependent EconomicActivity, y

As noted in the introduction, for mostcountries, mangrove loss is associated with keyprimary sector economic development activ-ities, such as shrimp aquaculture, overhar-vesting of forests, mining, salt production,and agricultural expansion in coastal areas.From the Food and Agricultural Organiza-tion of the United Nations (FAO) (2000) anestimate was obtained of total aquacultureproduction in metric tons of all shrimp andprawn species for the same year of estimatedmangrove area for each country reported inTable 1. The authors average total shrimpaquaculture production over the total coast-line length of a country.5 To represent pro-duction, or development, of other key primaryeconomic activities that might cause man-grove loss we employed data on agriculturalGDP per head of population engaged inagriculture, from World Bank (1998) and FAO(1997). However, mangrove deforestation maynot just be related to certain primary sectoreconomic activities but also be affected bythe general level of economic development aswell as the pace of economic growth in acountry. To examine this hypothesis, theauthors also included GDP per capita andGDP annual growth (from World Bank,1998) in the analysis.

B. Accessibility of Mangrove Areas, a

The accessibility of mangrove areas avail-able for conversion may be affected by suchfactors as roads, coastal infrastructure, andthe concentration of populations in coastalareas. As an indicator of the state of develop-ment of a country's road network, the articleemployed data on the percentage of pavedroads (World Bank, 1998). For some countries,data, were obtained on the population in

coastal urban agglomerations for 1980 andprojections for the year 2000 (WRI, 1994).

C. Ecological and Coastal Conditions, b

The abundance of mangroves found in anycountry will depend on a number of ecologicaland coastal conditions, including length ofcoastline, rainfall, tidal conditions, tempera-ture, water quality, and so forth. Estimatesof the total length of coastline (in km) for 88of the countries listed in Table 1 were obtainedfrom WRI (1994) and National GeographicSociety (1981).6 Data for only a limited numberof countries on average annual rainfall, tem-perature, and tidal range are available fromSpalding et al. (1997).

D. Environmental Protection, g

The theoretical model suggests that theremaining mangrove area found in a countryis likely to increase with the amount of envir-onmental protection. Spalding et al. (1997) alsoprovide an estimate of the total number of pro-tected areas for each of the countries listed inTable 1. The authors employed this variable asthe indicator of the degree of environmentalprotection in each country.

E. Institutional Factors, z

As noted, recent analyses have focused onhow a variety of institutional factors, suchas land use conflict, security of ownership, orproperty rights, political stability, and the ruleof law, may affect deforestation. The cross-country database on institutional factors thatmost closely matches the range of countries andthe corresponding years for mangrove areaestimates indicated in Table 1 is from Banks(1990). The Banks data set contains data on thenumber of major Cabinet changes (1985 only),constitutional changes (1985 only), govern-ment coups (1985 only), political purges (1985and 1990), government crises (1985 and 1990),guerrilla warfare incidents (1985 and 1990),industrial strikes (1985 and 1990), politicalassassinations (1985 and 1990), and politicalrevolts (1985 and 1990). The Banks data alsoinclude an index of party fractionalization(1985 only). All these institutional variables

5. As noted in the introduction, extensive shrimp andprawn aquaculture production in coastal areas is mostlikely to be responsible for conversion of mangroves. If acountry's shrimp aquaculture production is small relativeto its coastline, there is likely to be less impact on the totalmangrove area of that country. 6. The one exception is Micronesia.


essentially reflect the degree of political stabi-lity in a country. Rather than include each vari-able separately, the authors combined theminto three different single indicators of overallpolitical stability for each country:� A political stability index (PSI85)

based on the variables for 1985 only (Cabinetchanges, constitutional changes, and govern-ment coups plus the party fractionalizationindex).7

� A political stability index (PSI8590)based on a composite of the 1985 and 1990observations for the relevant variables (purges,crises, guerrilla warfare, strikes, assassinations,revolts, and riots).8

� A political stability index (PSI ) based onthe average of the above two indices.All three indices range from zero (highest poli-tical stability) to one (lowest political stability).

F. Other Economic and DemographicVariables

In addition to the various economic, ecolo-gical and economic factors identified in themodel of section II that may determine thesize of mangrove area found in a country,other demographic and economic influencesmay also affect mangrove conversion acrosscountries. To represent these possible exogen-ous influences, the authors extended the dataset for estimating (9) to include several addi-tional variables. These were population den-sity, rural population density, the percentage ofthe total labor force in agriculture, agriculturalraw material exports as a percentage of totalmerchandise exports, total debt service (as a

percentage of exports), and the real rate ofinterest. The source of all these variables wasWorld Bank (1998).

G. Estimation Approach

Given that the complete cross-country dataset comprises observations for a single yearonly, estimation of the empirical relationship(9) across countries was conducted usingordinary least squares. The estimations wereperformed on the entire sample of all coun-tries in Table 1, a subsample of countries withper capita income less than $7,000 per per-son, and a subsample of countries with man-grove area greater than 25 km2. In allregressions, the Breusch-Pagan chi-squaredtest indicated the presence of heteroscedasti-city, which was adjusted through usingWhite's (1978) robust correction of the cov-ariance matrix. All regressions were alsotested for the appropriate functional formÐlinear, semi-log, and logarithmic. In everycase the linear functional form was preferred.Finally, the independent variables were alsoexamined for multicollinearity, which wasrejected in all cases. This was particularlyimportant in ensuring that total shrimp aqua-culture production per length of coastline,agricultural GDP per head of populationengaged in agriculture, GDP per capita,GDP growth, and total coastline lengthcould be included together as independentvariables in the estimation of equation (9).

Table 2 lists the subset of variables fromthe full data set that were employed in thefinal cross-country regressions of the factorsdetermining the remaining mangrove area ofa country. Some of the variables in the com-plete data set, such as population in coastalurban agglomerations projected for 2000,tidal range, total debt service, and the realrate of interest, proved to have too few obser-vations across countries; thus, their inclusionreduced the sample sizes of the regressions sig-nificantly. In addition, none of these variableswere found to be significant in the latter regres-sions, and the overall explanatory power ofthese estimations was poor. Two other ecolo-gical variables, average annual rainfall andtemperature, plus the remaining exogenousdemographic and economic variables (i.e.,population density, rural population density,the percentage of the total labor force inagriculture, and agricultural raw material

7. The number of Cabinet changes, constitutionalchanges, and coups were each converted into aweighted-index variable using the formula Xi/Max jX j,where Xi is the observation for the ith country and MaxjX j is the maximum observation in the cross-countrysample. The three resulting weighted-index variables,along with the party fractionalization index, were thenaveraged into the single political stability index variable,using the standard formula

PXij/Nj where Xij is the index

value of the jth political variable for the ith country andNj is the number of political variables.

8. The composite 1985 and 1990 data set for each ofthe political variables (purges, crises, guerrilla warfare,strikes, assassinations, revolts, and riots) was created byselecting the observation from either 1985 or 1990,depending on which year more closely corresponded tothe single-year estimate of mangrove area for each coun-try in Table 1. The resulting data set for each variablewas then converted into a weighted index using the for-mula Xi/Max jXj, and then all the indexed variables wereaveraged into the single political stability index variable,using the standard formula

PXij/N.


exports as a percentage of total merchandiseexports), were not significant in any regres-sions. In most cases the inclusion of one ormore of these variables distorted the estimationresults considerably as well as reduced the over-all explanatory power of the regressions. Withthe exception of average annual rainfall, noneof these variables were included in the finalestimations of equation (9), and thus theyare not listed in Table 2. Finally, of the threeindices of political stability, PSI8590 per-formed consistently better in most regressions.This variable was therefore the preferred indi-cator of political stability.

The 89 countries in Table 1 are drawnfrom five distinct global geographical zones(Spalding et al., 1997). These are East Asia,Southeast Asia, and South Asia (zone 1);Oceania (zone 2); Latin America and theCaribbean (zone 3); Western (i.e., AtlanticCoast) Africa (zone 4); and Eastern (i.e., IndianOcean and Red Sea Coasts) Africa and thePersian Gulf (zone 5). Countries located inzone 1 account for approximately 43% of thetotal mangrove area of all countries listed inTable 1, and countries from zone 3 representover 31% of total mangrove area. Given thatmost of the world's remaining mangrove areasare largely found in these two zones, the articleincludes dummy variables for countries from

zone 1 and zone 3, respectively, in the cross-country regressions (see Table 2). The presenceof dummy variables for these zones allows theauthors to test the hypothesis that the mean(expected) values of remaining mangrovearea in zones 1 and 3 are greater than inother geographical regions (i.e., zones 2, 4,and 5).

Finally, given that the data range for man-grove area estimates by country is from 1960to 1996, alternative dummy variables for coun-try observations from the earlier years of thesample were also employed. These includeddummies for observations for years before1975, 1981, and 1985. In addition, a year indi-cator variable was also included in alternativeversions of the regressions. However, none ofthe dummy variables or the year indicator vari-able proved significant, nor did they improvethe explanatory power of the regressions.These variables were therefore dropped fromthe estimations.

IV. ESTIMATION RESULTS

The cross-country regressions of equation(9) for the sample of all countries are reportedin Table 3, along with the relevant test statis-tics. Four different versions of the regres-sion are indicated. Model 1 excludes both

TABLE 2

Definitions of Variables Used in Cross-Country Regressions

Variable Definition

MANGROVE Mangrove area (km2), year varies (Spalding et al., 1997).

AREAS Number of protected areas, same year as MANGROVE (Spalding et al., 1997)

AQUAS Total shrimp and prawn aquaculture production (metric tons),same year as MANGROVE, per length of coastline (km) (FAO, 2002;WRI, 1994)

AGDPAP Agriculture value added divided by population engaged in agriculture(constant 1987 US$/person), same year as MANGROVE(FAO, 1997; World Bank, 1998)

GDPG GDP growth (annual %), same year as MANGROVE (World Bank, 1998)

GDPPC GDP per capita (constant 1987 US$/person), same year as MANGROVE(World Bank, 1998)

ROADS Percentage of paved roads, same year as MANGROVE (World Bank, 1998)

COAST Length of coastline (km) (WRI, 1994).

CUP1980 Population in coastal urban agglomerations (thousands) (WRI, 1994)

PSI8590 Political stability indexa

RAIN Mean monthly annual rainfall (mm) (Spalding et al., 1987).

DUM1 Dummy for countries from zone 1 (Asia)b

DUM3 Dummy for countries from zone 3 (Latin America and Caribbean)

aConstructed based on Banks (1990); see text for explanation.bIncludes countries from East Asia, Southeast Asia, and South Asia.


the GDP per capita and growth variables.Model 2 includes GDP growth only, andmodel 3 includes GDP per capita only. Thefourth model displayed in Table 3 includesan additional ecological variable (RAIN) tomodel 3.9

Model 4 is the preferred regression. Theparameter estimate for GDP per capita(GDPPC) in the latter regression is significantand positive. Although average annual rainfallis significant only at the 10% level, it is posi-tively correlated with mangrove area acrosscountries. In addition the explanatory powerof model 4, as measured by the adjusted R2, isthe highest for all four regressions.

TABLE 3

Cross-Country Regression of Mangrove Area, All Countries (Dependent variable:MANGROVE [Mangrove area, km2])

Variables Model 1 Model 2 Model 3 Model 4#

AREAS 4.233� 101

(1.292)4.210� 101

(1.282)4.956� 101

(1.511)5.371� 101

(1.712)y

AQUAS ÿ2.611� 101

(ÿ1.595)ÿ2.531� 101

(ÿ1.534)ÿ3.163� 101

(ÿ1.996)*ÿ3.577� 101

(ÿ2.066)*

AGDPAP ÿ1.186� 10ÿ3

(ÿ3.261)**ÿ1.184� 10ÿ3

(ÿ3.246)**ÿ1.511� 10ÿ3

(ÿ3.255)**ÿ1.651� 10ÿ3

(ÿ3.509)**

GDPG ÿ1.729� 101

(ÿ0.420)GDPPC 2.209� 10ÿ1

(1.940)y2.962� 10ÿ1

(2.602)**

ROADS 1.031� 101

(0.682)9.659

(0.640)ÿ1.494

(ÿ0.103)ÿ1.881

(ÿ0.126)

COAST 7.170� 10ÿ1

(8.242)**7.188� 10ÿ1

(8.215)**7.097� 10ÿ1

(8.378)**7.104� 10ÿ1

(9.013)**

RAIN 4.049� 10ÿ1

(1.918)y

CUP1980 ÿ5.199� 10ÿ2

(ÿ1.291)ÿ5.276� 10ÿ2

(ÿ1.307)ÿ5.735� 10ÿ2

(ÿ1.501)ÿ5.769� 10ÿ2

(ÿ1.576)

PSI8590 ÿ7.851� 103

(ÿ3.196)**ÿ8.022� 103

(ÿ3.214)**ÿ7.059� 103

(ÿ2.809)**ÿ5.830� 103

(ÿ2.001)*

DUM1 2.021� 103

(1.773)y2.094� 103

(1.834)y2.491� 103

(1.929)y2.218� 103

(1.798)y

DUM3 2.077� 103

(3.495)**2.061� 103

(3.465)**2.032� 103

(3.365)**1.665� 103

(2.682)**

Constant 7.392� 102

(1.402)8.044� 102

(1.397)5.107� 102

(0.927)ÿ2.878� 102

(ÿ0.373)

N� 54 N� 54 N� 51 N� 49

F [9,44]� 44.12** F [10,43]� 38.89** F [10,40]� 38.38** F [11,37]� 45.85**

adj. R2� 0.880 adj. R2� 0.877 adj. R2� 0.882 adj. R2� 0.886

B-P c2� 15.17** B-P c2� 15.87** B-P c2� 13.63** B-P c2� 14.63**

*Significant at 5% level.ySignificant at 10% level.

**Significant at 1% level.

Notes: t-ratios are indicated in parentheses. N�number of observations. F [i, j ]�F-test of model. adj R2�adjusted R2 goodness of fit measure. B-P c2�Breusch-Pagan chi-squared test for heteroscedasticity. #Preferredregression.

9. Other variants than the four models depicted inTable 3 were also estimated. For example, includingboth GDP per capita and GDP growth did not improveon model 3, and the coefficient for GDPG was insignificant.Including a GDP per capita squared term in model 3 did notimprove the regression results, and the coefficient for the

squared GDPPC term was also not significant. Thus anenvironmental Kuznets curve income effect can be rejectedin this analysis.


In all four models the estimated coefficientsfor agricultural GDP per person employed inagriculture (AGDPAP) are highly significantand negative. In the preferred regression(model 4) as well as in model 3, the parameterestimate for shrimp aquaculture production(AQUAS) is also significant and negative.Thus the hypothesis that an increase in primarysector activities reduces the remaining man-grove area of a country cannot be rejected.

Model 2 indicates that GDP growth(GDPG) is not a significant variable in theregressions. Thus, mangrove deforestationdoes not appear to be affected by a country'srate of economic growth. In contrast, in thepreferred regression of model 4, GDPPC is sig-nificant and positive. This implies that thehypothesis that the general level of economicdevelopment (i.e., GDP per capita) in a countryinfluences mangrove deforestation cannot berejected. Across the sample of all countries,an increase in GDP per capita tends to be asso-ciated with more remaining mangrove areasin a country. The most reasonable explana-tion is that as an economy develops, manufac-turing and tertiary economic activities aremore prevalent, and these sectors are less likelyto be responsible for widespread mangrovedeforestation.10

Finally, the dummy variable for zone 3 wassignificant and positive in all four regressionsof Table 3, and the dummy variable for zone 1was positive and significant at the 10% level.Thus, the hypothesis that the mean values ofremaining mangrove areas in zone 3 (LatinAmerica and the Caribbean), and possibly inzone 1 (East Asia, Southeast Asia, and SouthAsia), are greater than in other geographicalregions cannot be rejected. Of the remainingvariables, the length of coastline and the poli-tical stability index were highly significant andpositive across all four estimations, whereas thepercentage of paved roads, the number of pro-tected areas, and coastal urban population in1980 were not significant.

Table 4 reports the two best regressions ofequation (9) with respect to each of the twosubsamples of countries, those with per capitaincome less than $7,000 and countries with

mangrove area greater than 25 km2. Models 1and 2 in Table 4 are for low- and middle-incomecountries (i.e., with GDP per capita less than$7,000 per person). Models 3 and 4 in Table 4are for countries with relatively large remainingmangrove areas (i.e., with mangrove areasgreater than 25 km2).

Neither GDP per capita nor GDP growthproved significant in any of the regressions forlow- and middle-income countries, and theoverall explanatory power of the estimationsthat included the latter variables was poor.Thus, GDP per capita and GDP growth areexcluded in the regressions reported in Table 4.The difference between the two reported esti-mations is that model 2 includes RAIN,whereas model 1 does not. However, RAINappears not to be significant in model 2, somodel 1 is the preferred regression.

In both model 1 and model 2 the estimatedcoefficients for both AQUAS and AGDPAPare negative but not significant. Thus, forlow- and middle-income countries the hypoth-esis that an increase in primary sector activitiesreduces the remaining mangrove area of acountry can be rejected.

In sum, models 1 and 2 suggest that neitherprimary production nor overall economicdevelopment and growth appear to be signifi-cant in influencing the remaining mangroveareas across all low- and middle-income coun-tries. In contrast, the key variables affectingmangrove areas for this group of countriesappear to be the length of coastline, the numberof protected areas, coastal urban population,political stability, and the percentage of pavedroads. The coefficients of these variables allhave the same signs as those predicted by thetheoretical model. Coastline length is onceagain positively associated with remainingmangrove area. Mangrove areas increase withenvironmental protection but decline as theseforests become more accessible due to animprovedroadnetwork,ascoastalurbanpopu-lations rise, or as political instability increasesin low- and middle-income economies.

Finally, the zonal dummy variables were notsignificant in either models 1 or 2 of Table 4,indicating that for low and middle-incomecountries there is no geographical differenceacross the five zones with respect to the regres-sions of remaining mangrove area.

As noted above, models 3 and 4 in Table 4are the best regressions for countries withrelatively large remaining mangrove areas

10. This finding and explanation is consistent withsome cross-country studies of deforestation, especiallyfor tropical countries, which also report a negative relation-ship between forest loss and GDP per capita. For recentreviews see Kaimowitz and Angelsen (1998) and Barbierand Burgess (2001).


(i.e., with mangrove areas greater than 25 km2).In both regressions, GDP per capita is signifi-cant and is included.11 The main differencebetween the two reported regressions is thatmodel 4 includes RAIN, whereas model 3excludes this ecological variable. However,RAIN is not significant in model 4, and somodel 3 is the preferred regression.

In both models the estimated coefficients forboth AQUAS and AGDPAP are significantand negative. Thus, for economies with man-grove areas greater than 25 km2, the hypothesisthat an increase in primary sector activitiesreduces the remaining mangrove area of acountry cannot be rejected. The parameter esti-mate for GDPPC in models 3 and 4 is alsosignificant and positive. For countries withlarge mangrove areas, one cannot reject thehypothesis that the general level of economicdevelopment (i.e., GDP per capita) affectsmangrove deforestation. Moreover, an

TABLE 4

Cross-Country Regression of Mangrove Area, Selected Countries (Dependent variable:MANGROVE [Mangrove area, km2])

Countries with Per Capita IncomeLess than $7,000 per Person

Countries with Mangrove AreaGreater than 25 km2

Variables Model 1# Model 2 Model 3# Model 4

AREAS 1.503� 102 1.515� 102 7.996� 101 8.067� 101

(5.763)** (5.808)** (5.598)** (5.891)**

AQUAS ÿ1.430� 101 ÿ1.806� 101 ÿ2.934� 101 ÿ3.386� 101

(ÿ1.085) (ÿ1.276) (ÿ2.405)* (ÿ2.463)*

AGDPAP ÿ8.622� 10ÿ4 ÿ1.030� 10ÿ3 ÿ2.228� 10ÿ3 ÿ2.259� 10ÿ3

(ÿ1.319) (ÿ1.604) (ÿ7.490)** (ÿ8.087)**

GDPPC 3.517� 10ÿ1 3.692� 10ÿ1

(2.139)* (2.410)*

ROADS ÿ2.683� 101 ÿ2.256� 101 ÿ2.589� 101 ÿ1.940� 101

(ÿ2.447)* (ÿ1.995)* (ÿ1.822)y (ÿ1.288)

COAST 6.994� 10ÿ1 7.012� 10ÿ1 7.035� 10ÿ1 7.059� 10ÿ1

(32.76)** (33.19)** (19.54)** (20.29)**

RAIN 1.631� 10ÿ1 3.066� 10ÿ1

(0.827) (1.426)

CUP1980 ÿ1.099� 10ÿ1 ÿ1.168� 10ÿ1 ÿ1.016� 10ÿ2 ÿ1.304� 10ÿ2

(ÿ2.370)* (ÿ2.332)* (ÿ0.477) (ÿ0.616)

PSI8590 ÿ7.851� 103 ÿ4.908� 103 ÿ7.741� 103 ÿ7.067� 103

(ÿ2.322)* (ÿ1.936)* (ÿ3.019)** (ÿ2.361)*

DUM1 1.066� 103 9.368� 102 1.496� 103 1.269� 103

(1.198) (1.069) (1.352) (1.169)

DUM3 8.223� 102 9.044� 102 1.501� 103 1.450� 103

(1.304) (1.377) (2.691)** (2.528)*

Constant 1.505� 103 6.369� 102 1.190� 103 4.295� 102

(2.850)** (1.819)y (2.021)* (0.504)

N� 45 N� 44 N� 46 N� 45

F[9,35]� 99.38** F[10,33]� 85.62** F[10,35]� 59.72** F [11,33]� 53.26**

adj. R2� 0.953 adj. R2� 0.952 adj. R2� 0.929 adj. R2� 0.929

B-P c2� 15.23** B-P c2� 15.21** B-P c2� 18.82** B-P c2� 16.50**

*Significant at 5% level.ySignificant at 10% level.


Notes: t-ratios are indicated in parentheses. N�number of observations. F [i, j]�F-test of model. adj R2� adjustedR2 goodness of fit measure. B-P c2�Breusch-Pagan chi-squared test for heteroscedasticity. #Preferred regression.

11. Replacing GDP per capita with GDPG, or includ-ing both variables together, in models 3 and 4 did notimprove the explanatory power of the estimations, andthe coefficient for GDPG was always insignificant.


increase in GDP per capita tends to be asso-ciated with more remaining mangrove areas ina country, suggesting that as a country withlarge mangrove areas develops economicallyless deforestation occurs.

The dummy variable for zone 3 was signifi-cant and positive for both regressions for coun-tries with large mangrove areas. Thus, for thesecountries, the hypothesis that the mean valuesof remaining mangrove areas in Latin Americaand the Caribbean (zone 3) are greater than inother geographical regions cannot be rejected.Of the remaining variables, the number of pro-tected areas, the length of coastline and thepolitical stability index are significant in bothmodels 3 and 4, and the percentage of pavedroads is significant at the 10% level in the pre-ferred model 3. The signs are once again aspredicted by the theoretical model. For coun-tries with large mangrove areas, these areas aregreater for countries with longer coastlines andmore environmental protection but decline forcountries with increased political instabilityand easier forest access due to an improvedroad network.

Table 5 indicates the elasticity estimates forthe explanatory variables in the preferredregressions of mangrove area for all three sam-ples of countries: (1) the sample of all countries(Table 3, model 4); (2) the subsample of coun-tries with per capita income less than $7,000(Table 4, model 1); and (3) the subsample ofcountries with mangrove area greater than

25 km2 (Table 4, model 3). The elasticity esti-mates facilitate comparisons across these threeregressions.

Across the entire sample of countries andfor countries with large mangrove areas, theestimated elasticities for both AQUAS andAGDPAP are highly significant and negative.Increases in overall agricultural GDP per per-son employed in agriculture have the largerimpact on mangrove loss. A 10% rise in AGD-PAP is related to a 10.9% fall in mangrove areafor all countries and a 10.5% decline for coun-tries with more than 25 km2 of mangroves. Incomparison, a 10% increase in shrimp aquacul-ture production is associated with 0.18% and0.14% declines in the respective subsamples.

For all countries and for those with largemangrove areas, the level of economic devel-opment, as represented by the GDP per capitaofacountry,ispositivelyassociatedwithgreatermangrove area. A 10% increase in GDP percapita is related to a 6.5% rise in mangrovearea across all countries and a 4.9% increasefor countries with large mangrove areas.

Several explanatory variables also haveimpacts on remaining mangrove areas,although with the exception of length of coast-line and political instability, most variableswere not significant across the three regres-sions. However, in all cases the effects of theexplanatory variables were the same as pre-dicted by the theoretical model. The numberof protected areas, coastline length, and

TABLE 5

Estimated Elasticities of the Cross-Country Regressions of Mangrove Area

VariablesAll Countries(N� 89)

Countries with Per Capita Income Less than$7,000 (N� 76)

Countries with MangroveArea Greater than25 km2 (N� 67)

AREAS 0.181y 0.347** 0.251**

AQUAS ÿ0.018* ÿ0.007 ÿ0.014*

AGDPAP ÿ1.089** ÿ0.292 ÿ1.046**

GDPPC 0.647** Ð 0.486*

ROADS ÿ0.032 ÿ0.349* ÿ0.307y

COAST 1.104** 0.889** 1.011**

RAIN 0.351y Ð Ð

CUP1980 ÿ0.131 ÿ0.200* ÿ0.019

PSI8590 ÿ0.165* ÿ0.165* ÿ0.195**

*Significant at 5% level.y Significant at 10% level.


Notes: N� number of observations in full sample. Elasticities calculated using all the nonmissing observations of thefull sample for each variable.


rainfall had positive effects on mangrove area,whereas the percentage of paved roads, coastalurban population, and political instability hadnegative impacts.

A country with a 10% greater coastlinelength will have 11% more mangrove areaacross all countries, 8.9% more in low- andmiddle-income economies, and 10.1% morein countries with large mangrove areas. A10% increase in political instability is asso-ciated with a 1.7% decline in mangrove areaacross all countries and in low- and middle-income countries, and a 2% decline in countrieswith mangrove areas greater than 25 km2. A10% increase in environmental protection isassociated with a 3.5% increase in mangrovearea in low- and middle-income economiesand a 2.5% increase in countries with largemangrove areas. A 10% rise in the percentageof paved roads and coastal urban populationslead, respectively, to 3.5% and 2% falls in man-grove area in low- and middle-income eco-nomies. Finally, although the variable issignificant only at the 10% level, a countrywith 10% more rainfall is likely to have 3.5%more mangroves across all countries.

V. CONCLUSION

This article sought to analyze the roleof economic development, in particular pri-mary-sector activities dependent on mangroveconversion, in causing the rapid decline incoastal mangrove areas across many countriesof the world. From a theoretical model of eco-nomic activity and mangrove conversion,in which the demand for land by convertingactivities leads to mangrove loss, the authorsestablished a relationship between remainingmangrove area, the level of economic activity,and other important causative factors. Man-grove area is expected to be a decreasingfunction of aggregate economic activities,such as shrimp aquaculture and agriculturalexpansion, that may involve the conversionor depletion of mangroves and to be increasingwith the amount of environmental protection.Ecological and coastal conditions, the accessi-bility of mangrove areas and institutional fac-tors (such as political stability) should alsohave an influence.

The mangrove area relationship emergingfrom the theoretical model was then estimatedempirically through a cross-sectional analysisof 89 countries, as well as for subsamples of

middle- and low-income economies and coun-tries with mangrove areas greater than 25 km2.The main purpose of the empirical analysiswas to examine the hypotheses that man-grove deforestation may be (a) due to certainmangrove-dependent primary sector economicactivities (i.e., aquaculture production andagricultural GDP per person employed in agri-culture) and (b) affected by the general level ofeconomic development and/or current rate ofgrowth in a country (represented by GDP percapita and GDP annual growth, respectively).

With regard to the first hypothesis, theregressions across all samples of countries andforcountrieswithmangroveslargerthan25 km2

provide some evidence that mangrove loss isassociated with expansion of shrimp aquacul-ture production along coastlines and primarysector (i.e., agricultural) activities generally. Ofthese two factors associated with mangrovedeforestation, increases in overall agriculturalGDP per person employed in agriculture havethe larger impact on mangrove loss.

With regard to the second hypothesis, thereis some evidence that the level of economicdevelopment of a country affects the amountof its remaining mangroves. However, the rela-tionship between mangrove area and the gen-eral economic performance of a country variesconsiderably depending on the sample of coun-tries. For all countries and for countries withlarge mangrove areas, the level of economicdevelopment appears to be positively asso-ciated with greater mangrove area. GDP percapita may be associated with expanding man-ufacturing and tertiary economic activities,and these economic sectors are less likely tobe responsible for widespread mangrove defor-estation. However, GDP per capita is not sig-nificantly associated with mangrove loss in thesample of low- and middle-income countries,and economic growth does not appear to influ-ence mangrove area at all.

Length of coastline seems to have a con-sistent effect across regressions. As expected,a country with a greater coastline length willhave more mangrove area. There was someevidence that a country with greater rainfallwill also have more mangrove area, but thisvariable was significant only at the 10% leveland only for the regression sample for all coun-tries (model 4 in Table 3). One interestingfinding of the analysis was that political in-stability appears to be strongly associatedwith mangrove deforestation in all regressions.


Other explanatory factors also influencedthe mangrove area of a country, althoughnot necessarily for all versions and samplesof the regressions used to estimate equation (9).However, in all cases, the effects of these vari-ables were the same as predicted by our theo-retical model. For low- and middle-incomeeconomies and for countries with large man-grove areas, an increase in environmental pro-tection is associated with an increase inmangrove area, whereas a rise in the percentageof paved roads is linked to a decline inmangroves (see Table 5). For low- and middle-income countries an increase in coastal pop-ulation is also associated with mangrove loss.

In conclusion, this analysis suggests that cer-tain forms of economic development in somecountries are likely to be associated with wide-spread mangrove loss. The authors find someevidence that mangrove loss is associated withexpansion of shrimp aquaculture productionalong coastlines and primary sector (i.e., agri-cultural) activities generally, but not necessa-rily in all low- and middle-income countries.Instead, it appears that countries with largemangrove areas are particularly susceptibleto mangrove conversion from primary-sectoractivities. Much of the current literature onglobal mangrove deforestation has focusedon the influence of shrimp aquacultureexpansion, which appears to be particularlydevastating to mangroves in a handful ofmajor shrimp-exporting countries (Barbierand Cox, 2002; Primavera, 1997). In contrast,this analysis suggests that over a wide range ofcountries globally with remaining mangroveareas, increases in overall agricultural GDPper person employed in agriculture have thelarger impact on mangrove loss. Nevertheless,from a policy perspective, there seems to beconsiderable scope for countries to reducemangrove deforestation through mitigatingthe impacts of all primary sector activities incoastal areas, including shrimp aquaculture,on remaining mangrove areas.

The authors also find some evidence that asa mangrove-rich country develops economic-ally and presumably becomes less dependenton primary-sector activities, it may reduce itsmangrove deforestation. On the other hand, ifthecountry ispoliticallyunstableandhasacces-sible mangrove areas due to improved roadnetworks, these factors will also contribute toincreased mangrove loss. However, increasedinvestment in environmental protection will

counteract this loss somewhat. Perhaps thisis one encouraging policy option emergingfrom this analysis. If the international commu-nity can provide technical and financial assis-tance to developing economies with largeremaining mangrove areas to increase theirprotection and conservation efforts, then thismay be one way of slowing down rates of man-grove loss globally.

Finally, there are a number of importantcaveats to this analysis, given the limiteddata set. It would have been preferable tohave mangrove data for more than one yearacross countries, but unfortunately suchtime-series observations are not available formany countries. The absence of such time-ser-ies data limited the authors' ability to employmore sophisticated panel analyses of changes inmangrove area over time. A second problemwith the data set is that some key ecologicalvariables, such as temperature and tidal levels,were incomplete for many countries, and otherimportant ecological indicators, such as levelsof coastal pollution, are not available. Othervariables, such as rainfall, percentage of roadspaved, and number of protected areas, arenational aggregates rather than specific tocoastal areas where mangroves are located.Even key economic variables were not alwaysavailable for all the countries in the sample forthe years indicated in Table 1. As can be seen inTables 3 and 4, the result was that the samplesizes of regressions were often smaller than thefull sample of 89 countries.

Given these limitations, the analysis andconclusions of this article of the potentialimpacts of economic development on man-grove deforestation across countries must beconsidered preliminary. Nevertheless, becausethis article represents the first attempt todevelop a formal analysis of the economic fac-tors determining mangrove loss worldwide, theauthors hope that our effort will lead toimproved data collection and further analysison this important topic.

REFERENCES

Aksornkoae, S., S. Priebprom, A. Saraya, andJ. Kongsangchai. ` Mangrove Resources and theSocio-Economics of Dwellers in Mangrove Forestsin Thailand,'' in Man in the Mangroves, The Socio-Economic Situation of Human Settlements in Man-grove Forests, edited by P. Kunstadler, E. L. F. Bird,and S. Sabhasri.Bangkok:United NationsUniversityand the National Research Council of Thailand,1986.


Alston, L. J., G. D. Libecap, andB. Mueller. ` Land ReformPolicies, the Sources of Violent Conflict, and Implica-tions for Deforestation in the Brazilian Amazon.''Journal of Environmental Economics and Manage-ment, 39(2), 2000, 162±88.

Anderson, J. L., and S. W. Fong. ` Aquaculture and Inter-national Trade.'' Aquaculture Economics and Man-agement, 1(1), 1997, 29±44.

Banks, A. S. Cross-National Time-Series Data Archive.Centre for Social Analysis, State University ofNew York, Binghamton, 1990.

Barbier, E. B. ` Habitat±Fishery Linkages and MangroveLoss in Thailand.'' Contemporary Economic Policy,21(1), 2003, 59±77.

Barbier, E. B., and J. C. Burgess. ` The Economics of Tro-pical Deforestation.'' Journal of Economic Surveys,15(3), 2001, 413±32.

Barbier, E. B., and M. Cox. ` Economic and DemographicFactors Affecting Mangrove Loss in the Coastal Pro-vinces of Thailand, 1979±1996.'' Ambio, 31(4), 2002,351±57.

Barbier, E. B., and I. Strand. ` Valuing Mangrove±FisheryLinkages: A Case Study of Campeche, Mexico.''Environmental and Resource Economics, 12, 1998,151±66.

Deacon, R. T. ` Deforestation and the Rule of Law in aCross-section of Countries.'' Land Economics, 70(4),1994, 414±30.

ÐÐÐ. ` Deforestation and Ownership: Evidence fromHistorical Accounts and Contemporary Data.''Land Economics, 75(3), 1999, 341±59.

Jitsanguan, T., B. Sootsukon, and S. Tookwinas. ` Estima-tion of Environmental Costs from Shrimp Farming.''Report submitted to Office of Environmental Policyand Planning, Ministry of Science, Technology andEnvironment by the Department of Agriculturaland Resource Economic, Faculty of Economics,Kasetsart University, 1999.

Kaimowitz, D., and A. Angelsen. Economic Models ofTropical Deforestation: A Review. Bogor, Indonesia:Center for International Forestry Research, 1998.

Kelleher, G., C. Bleakley, and S. Wells. A GlobalRepresentative System of Marine Protected Areas,volume 1. Great Barrier Reef Marine Park Autho-rity, World Bank, and World Conservation Union,1995.

Kongkeo, H. ` Comparison of Intensive ShrimpFarming Systems in Indonesia, Philippines, Taiwanand Thailand.'' Aquaculture Research, 28, 1997,789±96.

Maltby, E. Waterlogged Wealth: Why Waste the World'sWet Places? London: Earthscan Publications, 1986.

Mitsch, W. J., and J. G. Gosselink. Wetlands, 2d ed. NewYork: Van Nostrand Reinhold, 1993.

Mooney, H. A., J. Lubchenco, R. Dirzo, and O. E. Sala.` Biodiversity and Ecosystem Functioning: Ecosys-tem Analyses,'' in Global Biodiversity Assessment.UNEP, 1995, 387±93.

National Geographic Society. National Geographic Atlasof the World, 5th ed. Washington, DC: NationalGeographic Society, 1981.

Naylor,R.L.,R.J.Goldburg,J.H.Primavera,N.Kautsky,M. C. M. Beveridge, J. Clay, C. Folke, J. Lubchenco,H. Mooney, and M. Troell. ` Effect of Aquaculture onWorld Fish Supplies.'' Nature, 405, 2000, 1017±23.

Primavera, J. H. ` Socio-Economic Impacts of Shrimp Cul-ture.'' Aquaculture Research, 28, 1997, 815±27.

Raha, S. K., and M. M. Alam. ` Shrimp Farming: A Profit-able Enterprise in South-Western part of Bangla-desh.'' Economic Affairs, 42(2), 1997, 96±99.

Ruitenbeek, H. J. ` Modelling Economy±Ecology Linkagesin Mangroves: Economic Evidence for PromotingConservation in Bintuni Bay, Indonesia.'' EcologicalEconomics, 10, 1994, 233±47.

Sathirathai, S., and E. B. Barbier. ` Valuing MangroveConservation in Southern Thailand.'' ContemporaryEconomic Policy, 19(2), 2001, 109±22.

Spalding, M., F. Blasco, and C. Field. WorldMangrove Atlas. International Society for MangroveEcology, World Conservation Monitoring Centre,1997.

Tokrisna, R. ` The Use of Economic Analysis in Support ofDevelopment and Investment Decision in Thai Aqua-culture: With Particular Reference to Marine ShrimpCulture.'' Report to the Food and Agriculture Orga-nization of the United Nations, 1998.

United Nations Food and Agricultural Organization(FAO). FAOSTAT Statistical Database 1997.Rome: FAO, 1997.

ÐÐÐFISHSTAT Plus Version 2.3. Rome: FAO, 2000.

World Bank. World Development Indicators, CD-ROM.Washington DC: World Bank, 1998.

World Conservation Monitoring Centre (WCMC). Biodi-versity Data Sourcebook. Cambridge: World Conser-vation Monitoring Centre, 1994.

White, H. ` A Heteroskedasticity Consistent CovarianceMatrix and a Direct Test for Heteroskedasticity.''Econometrica, 46, 1978, 817±38.

World Resources Institute (WRI). World Resources1994±5. New York: Oxford University Press,1994.

ÐÐÐ. World Resources 1996±7. New York: Oxford Uni-versity Press, 1996.


does economic development lead to mangrove loss? a cross-country analysis

Documents