An assessment of alternative methods of estimating the effect of the ivory trade ban on poaching...

Ecological Economics 30 (1999) 93–106

ANALYSIS

An assessment of alternative methods of estimating the effect ofthe ivory trade ban on poaching effort

Michael Burton *Agricultural and Resource Economics, Uni6ersity of Western Australia, Nedlands, Perth, WA 6907, Australia

Received 28 April 1997; received in revised form 29 September 1998; accepted 5 October 1998

Abstract

The ban on international trade in ivory introduced by the international community in 1989 was intended to protectthe declining stock of African elephant (Loxodonta africana), by providing a barrier between the range state and theprincipal consuming nations. If there is a residual trade, then the efficiency of the ban is determined by the declinein the price paid to poachers, and the responsiveness of poaching effort to changing prices. This paper concentrateson the latter issue, and provides a review of a number of alternative approaches to modelling poaching effort, withextensions. It is found that even low residual prices would induce a level of effort that would significantly depresselephant numbers. © 1999 Elsevier Science B.V. All rights reserved.

Keywords: Ivory; Poaching effort; Trade ban; Economics

1. Introduction

In December 1989 the Convention on Interna-tional Trade in Endangered Species (CITES)voted to place the African elephant (Loxodontaafricana) on Appendix I of endangered species,effectively banning commercial trade in all ele-phant products between the signatories of thetreaty. Some Southern African states took a reser-vation, which would allow them to trade with

non-signatories or between themselves but theimplementation of the ban by domestic legislationwithin the consuming nations (principally theUSA, Japan and the European Union) and manyof the range states was an attempt to solve at theinternational level an environmental managementproblem using a trade instrument. The problemwas a rapidly declining African elephant popula-tion, which was estimated to have declined bysome 50% at a continental level between 1979 and1989 (with a point estimate of 609000 in 1989)and even more in some individual countries. Esti-mates for 1991 give a range of between 549000and 652000, although

* Fax: +61 8 9380 1098; e-mail: [email protected]

0921-8009/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved.

PII: S0921-8009(98)00102-5

M. Burton / Ecological Economics 30 (1999) 93–10694

‘‘..these figures should be viewed with extremecaution, as elephant numbers for a number ofcountries are based on informed guessworkonly.’’ (Douglas-Hamilton et al., 1992, p. 15)

The cause of this decline, particularly in the last20 years, was seen to be the illegal exploitation ofelephant for its ivory (although the decline overthe longer run has been attributed to direct com-petition for habitat between man and elephant:see Parker and Graham, 1989a,b,c). It is temptingto suggest that a review of current populationestimates should reveal the impact of the ban.Dublin et al. (1995) give some country level popu-lation estimates which can be compared with pre-ban estimates, and these give a very mixed pictureof increases and declines (Table 1). Overall, thereis a general picture of decline, apart from Zim-babwe, which has had a relatively well fundedwildlife protection system, and Kenya which overthe period has had a significant increase in fund-ing (Dublin et al., 1995, p. 23). However, suchcomparisons are fraught with difficulties, not leastthe problem of accurately estimating populations,and the effect of changes in other factors (e.g.enforcement expenditure, or land use policies).

The decision to place the African elephant onAppendix I was the culmination of a protracteddebate on their management. The SouthernAfrican states who held relatively large and stablepopulations wished to retain the right to trade,arguing that the revenues generated would help tofund conservation measures. Other African stateswhich had suffered extreme levels of poachingargued that by shutting off final demand theincentives for poaching would decline to a levelthat could be contained by law enforcement. In1997, at the tenth meeting of the conference of theparties to CITES, a limited return to trade inivory and elephant products was allowed, with thedownlisting of the elephant populations inBotswana, Namibia and Zimbabwe to AppendixII. Ivory quotas have been set for each countryfor 1998 and 1999, subject to the three rangestates and the proposed sole purchaser of theivory (Japan) meeting certain requirements relat-ing to monitoring of the trade, and using therevenues generated to support elephant conserva-

tion (Gray, 1997). The intention of this limitedchange is to allow those countries with stablepopulations and adequate management to realisesome economic return, while the controls on thetrade are intended to minimise the impact it mayhave on the other range states. The full implica-tions of the change have yet to be revealed.

The fundamental issues underpinning the over-exploitation (the lack of political will or economicmeans to enforce the property rights in the re-source) are not addressed by a ban, althoughdoubts as to its efficacy were raised in the IvoryTrade Review Groups report to CITES (IvoryTrade Review Group, 1989a,b) and by others(Barbier et al., 1990; Sugg and Kreuter, 1994;Kreuter and Simmons, 1995; Bulte and vanKooten, 1996). A polemical account of the politi-cal pressures and lobbying that led to the actualdecision is given in Bonner (1993) but there hasbeen little analysis of the relative economic effi-ciency of the ban as compared to other manage-ment options, although Barnes (1996) gives anassessment of the impact of the ban on economicvalues within Botswana, and Khanna and Har-ford (1996) indicate, within a theoretical frame-work, why a trade ban may have to besupplemented by economic incentives if it is to beeffective.

An assessment of a (totally effective) ban wouldrequire the comparison of the economic lossesassociated with adopting a second best interven-tion (i.e. at the consumption level to address theeffective lack of property rights at the producer

Table 1Elephant population estimates for selected countries, pre- andpost-ban

Post-banPre-ban

Zimbabwe 52 000 61 51523 797Kenya 16 000

Tanzania 61 000 54 15717 000Camaroon 22 000

2000Malawi 280032 000Zambia 22 785

Nigeria 10001300

Source: Pre-ban, Ivory Trade Review Group (1989a,b). Post-ban, Dublin et al. (1995).

M. Burton / Ecological Economics 30 (1999) 93–106 95

level) with the transactions costs associated withenforcing a trade ban as compared to those forenforcing the property rights in the range states.The latter can be extremely high, with estimates inthe range of $US 500/km2 in 1980 values (Parkerand Graham, 1989a; Burton, 1994a) so potentiallythe transactions costs of monitoring trade mayoutweigh the deadweight losses associated with aban.

We do not attempt such an assessment here,but look at a single aspect of the issue: an evalua-tion of the likely impact of the ban on the activi-ties of elephant poachers. The increase inmortality in the past 20 years has been linked tothe concurrent increase in the real price of ivory,but there have been few studies that have formallyestablished that link, nor identified its strength,despite its importance for the pro-ban argument.

In attempting to identify the factors affectingpoaching activity there are substantial data prob-lems. If one uses the definition of ‘effort’ commonin fisheries economics as an index of factor inputs,there is obviously no direct measurement ofpoachers’ effort, nor any component of it. Thereare some indicators of poachers’ output, namelythe carcass of poached elephants that may befound (minus tusks), but these do not measureeffort per se as one needs to correct for theelephant population being hunted, itself subject tolarge estimation errors. One also needs to accountfor other changes in the effectiveness of effort,such as changes in anti-poaching activity. Theempirical sections of this paper review and extendvarious studies which have investigated the factorsthat determine the level of poaching effort, whichis a necessary precursor to any assessment of theeffects of reduced final demand caused by a tradeban.

2. Direct estimation of poaching effort

This section reviews possible measures ofpoaching effort which can be inferred directly.The difficulty with this approach is the lack ofgood data on many of the aspects of the problem(Leader-Williams, 1993). Dublin and Jachmann(1992) report data for five African range states,

for a number of national parks in each country.The measure of poaching activity is taken as thecarcass ratio (defined as ratio of carcasses to liveelephants in an area). Variation in this ratioacross parks within a country (estimated for theperiod 1987–1989) is explained by elephant densi-ties and measures of anti-poaching activity (suchas budget/km2, scout/km2 etc.), using a linearregression based on pre-ban data. Lack of timeseries data means that there is only one value foreach park, so that the regression is based on fourto seven data points only, depending on the coun-try. The estimated equation and post-ban esti-mates of the explanatory variables are then usedto generate post-ban forecasts of carcass ratios,which are compared to the actual values (for1990–1991) using a Wilcoxon matched-pairsrank-sign test. Essentially they are testing forstructural change in the relationship betweenpoaching rates and enforcement with the intentionof ascribing any such break to a change in theincentive structure caused by the ban. In fact nosuch structural break is identified. In all cases thepost-ban poaching rate is consistent with a stablepoaching relationship across the time period whenthe ban was introduced. This negative assessmentof the impact of the ban has to be tempered bythe very low power of the tests used, due to thesmall number of observations. In a follow-upstudy (Dublin et al., 1995) an alternative mod-elling methodology is adopted, but with broadlythe same objective. Here the dependant variable isthe number of carcasses found at sites (usuallydefined as a national Park or similar area) withina country, for each year in the period 1988–1993.It is hypothesised that the number of carcasseswill be predicted by the area of the site, theenforcement budget, the number of scouts em-ployed in the area, the elephant population and adummy variable that takes a value of zero pre-ban, and one post-ban. The model is estimated foreach country separately. The inclusion of the‘ban’ variable allows for a direct statistical test ofthe impact of the ban: if it is not significant then,again, the variation in the carcass numbers isexplained solely by the other variables. Althoughthe number of countries surveyed is increased (tonine), in only three were the data of sufficient


Fig. 1. Carcases and number of incursions for regions of Zimbabwe, 1984–1992.

quality to allow the model to be estimated, and innone of these is the ban variable statisticallysignificant. Although this (negative) result has tobe qualified by the poor quality and quantity ofthe available data, their overall conclusion fromthis and the more general descriptive data col-lected (e.g. comparing carcass ratios pre- andpost-ban) is that the international ivory trade banhas not halted the illegal offtake of elephants;that, of those considered, only Zambia has hadany marked decline in carcass numbers; and thatmarkets for illegal ivory continue to flourish.

Given the low power of the tests used in thesestudies, an obvious extension is to conduct thetest for structural change using a longer timeseries of data. Data exists for Zimbabwe at a parklevel for a wide range of data on poaching activ-ity, including elephants shot, poachers arrested,incursions, etc. for the period 1980–1992. There isnot good data on anti-poaching activity at thepark level, although there is some at regionallevel. However, the author could find no relation-

ship between levels of poaching activity (howeverdefined) and enforcement. This may be because ofthe degree of aggregation of the data, but Fig. 1,which reports incursions in four regions, alsoindicates a spatial aspect to exploitation. Prior to1984 there are relatively few incursions, except inGonarezhou. There is then a sequencing of incur-sions, starting in the Zambezi region and movingthrough Hwange and Sebungwe, and this patternis despite a relative constancy in wildlife popula-tions and anti-poaching operations prior to thestart of poaching activities. What appears to oc-cur is not a simultaneous exploitation in all areas,but the exploitation of the ‘low-cost’ areas first(on the borders in the north and south east of thecountry) and as populations decline, effort has tobe extended into areas with higher costs, or lowerinitial population densities. Thus the effort ex-pended at the park level depends not only on themarginal returns to poaching in that park, butalso returns elsewhere. Making inferences aboutthe determinants of effort based on regional or


even park level data is then very problematic. Ithas been suggested that the process of sequentialexploitation of populations has occurred at acontinental level, with exploitation moving fromcountry to country as marginal returns decline,and the general increase in poaching in Zimbabweas a whole in the mid-1980s was not due tochanges in the internal position at the park level,but because of the total collapse in elephant popu-lations in East Africa (Martin, R., personal com-munication, 1991).

An analysis of the Zimbabwe data highlightstwo other issues. The first is multi-species ex-ploitation by poachers: associated with the in-crease in incursions in Fig. 1 is an initial increasein poaching of rhino, and only as rhino numbersdecline does elephant poaching increase, suggest-ing a further, local-level, evaluation of relativereturns. Secondly, there may be an apparent per-verse relationship between observed poaching ef-fort and enforcement. The immediate victims ofpoaching do not report the crime, this will gener-ally be identified by anti-poaching patrols, eitheron the ground or from the air. As one increasesenforcement one may record an increase in illegalactivity, even though actual illegal activity is de-clining. At some point the effect of increasedenforcement on activity will overwhelm the effecton detection, and the expected negative relation-ship will emerge. This problem is less likely toemerge if the periodicity is short, so that one canexpress poaching effort as a function of laggedenforcement as poachers may reduce effort asthey experience previous high levels of enforce-ment, but it places high demands on datacollection.

Jachmann and Billiouw (1997) undertake ananalysis based on a much more complete data set,derived from the Luangwa Integrated ResourceDevelopment Project in Zambia, which has main-tained detailed information on law-enforcementand illegal activities for the period 1988–1995.They explain the annual variation in the numberof elephant illegally killed in the area as a func-tion of the level of manpower and budget avail-able, measured using nine different variables.Because of the very small number of observations(eight), the analysis was restricted to using only

two explanatory variables in any one model: thenumber of bonuses paid and each of the othereight available explanatory variables, giving eightpossible models. Bonuses are paid if members of apatrol make an arrest, or confiscate a firearm ortrophy, or if information is provided which leadsto such outcomes. As such it reflects the level ofoverall enforcement activity, and is the variablewhich is most highly correlated with illegal off-take. Only two of the models are deemed ade-quate, on the basis of the overall significance ofthe equations: those that augment bonuses withthe number of effective investigation days (i.e.days spent investigating reports of illegal activity,usually in villages and towns rather than on pa-trol in the study area), and the number of carriersemployed to assist scouts when on patrol. How-ever, the latter was negatively related, which iscounter-intuitive, and the authors use the formerspecification (i.e. bonuses paid and effective inves-tigation days, both of which are negatively corre-lated with kills) in estimating the budgetary costof achieving acceptable levels of elephant losses.

There are two comments that should be madeabout the results. Firstly, although the number ofbonuses paid ‘‘incorporates the influence of thenumber of scouts and carriers employed as well asthe number of effective patrol days (i.e. the entiremanpower input).’’ (p. 240) it must also reflect thepropensity to poach. Thus one could imagine asituation where the incentives to poach decreasebecause of a change in the external conditions,which leads to a decrease in bonuses paid (asthere is less activity to counter) while at the sametime illegal killings decrease. The extent to whichthis endogeneity may be affecting the results isunclear. Secondly, although the data period in-cludes both pre- and post-ban periods, there is noformal attempt to test if this may be statisticallysignificant. Instead, on the basis of the overallsignificance of their preferred equation, they con-clude that ‘‘..the variation observed in the num-bers of elephants found killed illegally between1988 and 1995 can be explained by resource allo-cation, without any contribution of externalfactors including the international ivory ban.’’(p. 243). They therefore discount the possibilitythat allowing the ban to have some impact may


potentially increase the extent to which that varia-tion can be explained.

Direct estimation of the responsiveness ofpoaching activity to economic incentives and en-forcement activity is fraught with difficulties, bothin terms of data and methodology. It has to besaid that none of the data reviewed above give agreat deal of guidance on the impacts of the banon poaching effort.

3. Micro-simulation models

An alternative approach has been to developmicro-simulation models of poaching behaviour,based on detailed data collected at a single pointin time. An example of this is given in Milner-Gulland and Leader-Williams (1992) (hereafterMGLW), where various models of poaching ac-tivity are constructed. Here we concentrate ontheir open access model of elephant hunting. Indi-vidual gangs are assumed to maximise short-runprofits subject to a production function, whichrelates harvest levels to levels of effort and densityof elephants, and a cost function, which not onlyincludes direct costs of effort, but also the ex-pected fine level given a probability of capture.Four models are developed, depending on theform of the production function (linear or non-linear) and the form of the penalty function (a flatfine if caught, or one that varies with amounts ofpoached ivory taken). The models can be repre-sented by:

maxx Profit=pq−wx−F

Subject to: a production function:

q=aBx (linear) or q=a %Bx0.5 (non− linear)

and a penalty (fine) function:

F=bx(r+p) (constant fine) or

F=bx(rq+p) (variable fine) (1)

where p is the price per unit of output; q is thequantity of output; x is the quantity of input(number of poaching trips); w is the price per unitof input; B is the elephant population; r is the finerate; F is the expected fine; and a,a %,b are thetechnical parameters.

The probability of being caught is given by bx,and in the variable fine case the fine is given by rtimes q, the output level, plus the value of oneconfiscated trophy (p); in the constant fine case itis simply r+p.

Using this model, it is possible to identify theoptimal level of input for the firm, x*, under thedifferent assumptions about the nature of theproduction and penalty functions. For the case ofthe variable fine and linear production functionthis is given by:

x*=paB−w−bp

2braBif pab−w−bp\0

x*=0 otherwise

(2)

MGLW report extensive data obtained in thefield from which the parameters and economicvariables can be derived (Milner-Gulland andLeader-Williams (1992), Appendix 1 and 2), andwhich allow one to translate this expression into anumerical estimate of poaching activity. Usingdata based on 1985 values (and in particular 1985elephant population levels) the optimum is 12expeditions, which generates 42.4 units of output.This gives an estimate of the hunting mortality(defined as the proportion of the population thatit is optimal to kill) of 0.003, as reported in Table1 of MGLW. This is the short-run firm-levelresponse. The industry-level short-run response issimply the aggregate of the (fixed number) ofindividual gangs. However, in the context of thecurrent paper, what is of more interest is thelong-run industry equilibrium response, and itsresponsiveness to the ivory price, and this is notdealt with in MGLW.

At the short-run level of exploitation each firmis generating profits, and in an open access systemone would expect to see additional firms enteringthe industry until all profits have been driven outof the system. The mechanism by which thisoccurs is falling population numbers: as morefirms enter and harvest their optimal 42.4 units,total population will fall. This will reduce theoptimal level of input x* and the optimal level ofoutput q* for each firm but the process of adjust-


ment will continue until an equilibrium level ofelephant numbers is identified. This will occurwhere the average profit earned by each firm iszero. This can be identified by substituting theexpression for the optimal level of input, Eq. (2),into the definition of profit (Eq. (3)):

p=paxB−wx−bx(raxB+p) (3)

Given that all gangs are assumed to be identi-cal, and there is no short-run interaction betweenthe firm-level profit and the number of firms, onecan identify the long-run equilibrium populationlevel by setting the profit function equal to zeroand solving for B, i.e. identifying the level ofpopulation at which firm-level profits will bedriven to zero when the firm uses the optimal levelof input (if one has identified all of the costscorrectly then firms will still be prepared to oper-ate at this level, as there will be no better alterna-tive use of their resources). The solution for Be inthis case is:

Be=bp+w

pa(4)

Thus, in principle, one can identify the level towhich the open access exploitation of the resourcewill drive the population level in its search forprofit.

However, in the case of a linear productionfunction and variable fine, the analysis runs intoconceptual difficulties. The system will only reachan equilibrium when the level of the elephantpopulation has been reduced to such a low levelthat the optimal level of input for the indi6idualgang is zero (this can be seen by substituting Eq.(4) into Eq. (2)). This occurs because at popula-tion levels higher than Be it is possible for firms toidentify a level of input that generates a profit forthe individual, and hence attract further effortinto the system, and so further reduce the popula-tion level.

This leads to the internally consistent result thatat the industry equilibrium there will be a positivelevel of output produced by an infinite number ofproducers, each producing an infinitely small levelof output. This is an analogue to the conventionallong-run analysis of a perfectly competitive indus-try where individual firms face decreasing returns

to scale at all output levels, and hence the mini-mum of the average cost curve occurs at aninfinitely small level of output. At prices abovethis level, firms enter in response to profit, in-creasing industry level output and hence pushingprices down towards the minimum average cost.In the current model, entry causes elephant num-bers to decline rather than the price of output,and hence costs to increase, but the basis of theanalysis is identical.

However, this analytical possibility assumesthat inputs are completely divisible. In the currentcase the input (expeditions) should strictly takeinteger values only. This will not lead to largeerrors if the number of expeditions is large, but atsmall numbers, and in particular at values close tounity, the constraint imposed by the indivisiblenature of the input has to be reflected in themodel solution. Thus, one can replace the opti-mality condition Eq. (2) for the level of input bythe corner solution of a single expedition pergang. One can follow the same logic as above,and identify the population level at which theprofit per gang falls to zero by setting x=1 in Eq.(2) above, and solving for Be:

Be=w+bp

pa−bra(5)

This gives an equilibrium elephant populationunder open access of 998, or a 93% mortality ratecompared to the 1985 population.

It is not possible to identify the aggregate levelof input in the industry needed to maintain thispopulation level, nor the aggregate output, with-out knowing the growth function of the elephant,but the form of the growth function does notaffect the equilibrium level of the populationwhich is determined solely by the profitability ofthe individual firms at different levels of popula-tion, not by the rate of growth of the population.

If one changes the form of the function govern-ing fines so that a constant fine is charged, theneven the firm-level analysis runs into difficulties.Because there is no decreasing marginal produc-tivity in the use of input, nor increasing marginalcost component, in the short-run, if the individualfirm finds it profitable to use a single unit of inputthen there is no limit as to the level of input that


will generate marginal profits. Thus the results ofthe short-run firm-level analysis is indeterminant.MGLW recognise this (p. 391) and suggest thatthe population size is varied until the profitearned in the industry is zero.

The equilibrium level of the population is givenby (Eq. (6)):

Be=w+bp+br

pa(6)

and this yields the hunting mortality of 91.9%reported in Table 1 of MGLW.

The model is also solved when the non-linearproduction function is used. For both fine struc-tures, for the single firm, there exists a positivelevel of input (however small) that will generate apositive level of profit ‘for any set of prices, costs,fines and population le6el’. This may seem counter-intuitive, but it derives from the form of theproduction function: as input tends to zero, themarginal productivity of the input tends to infi-nity. This means that as elephant numbers fall inresponse to effort entering the industry, the indi-vidual firms will adjust their input downwards butwill always be earning positive profits. This leadsto similar difficulties at the industry-level as be-fore: the search for profit will force populationlevels down as long as there are positive profits tobe earned, and this will only stop when the firm’sinputs have been driven to very small levels.

However, because of the indivisibility of theinput, it is not possible for input levels to fallbelow unity. Thus, again, one can identify thelong-run industry-level equilibrium level of popu-lation by identifying that level of population thatgenerates zero profits when one unit of input isused. For the variable fine case the definition ofprofit is (Eq. (7)):

p=pa %Bx0.5−wx−bx(ra %Bx0.5+p) (7)

which yields (Eq. (8)):

Be=w+bp

pa %−bra %(8)

which gives an equilibrium population of 258 (or98.1% mortality). Note that this differs from thepopulation derived from Eq. (5) because theparameters of the production function differ.

The constant fine case gives (Eq. (9)):

p=pa %Bx0.5−wx−bx(r+p) (9)

which yields:

Be=w+bp+br

pa %(10)

which gives an equilibrium population of 288 (or97.9% mortality).

The higher mortality rates when using the non-linear production function arise because theparameters used imply a higher level of productiv-ity at low levels of input compared to the linearfunction. What this analysis indicates is that usinga variable fine structure has little impact on theequilibrium population towards which the openaccess system will converge, whereas the represen-tation of the technology has a greater effect.

It is also possible to identify the responsivenessof the system to changes in the economic parame-ters by calculating the appropriate elasticities.These are reported in Table 2, for the case of thelinear production function only. The first columnreports the elasticity of output for the individualfirm, for the variable fine case, based on the 1985data. It indicates that the firm’s output is rela-tively responsive to the price of output and thelevel of the fine, but not to the level of the wage.At the firm-level poachers are also responsive tothe probability of detection. However, a differentpicture emerges at the industry-level. Column 2reports the elasticities of the ‘industry’ equilibriumpopulation level. These are measured close to theequilibrium population identified by Eqs. (4) and(5) (i.e. 1000 elephant). The signs are reversed ascompared with the firm-level analysis (here thedependent variable is the residual populationrather than the firm’s offtake), but what is notableis the change in the effect of the wage rate andfine level. These results are not inconsistent: onereports the responsiveness of an individual firm,the other the effect on the number of firms whichcan profitably operate in the system. The finalcolumn reports the elasticities for the equilibriumpopulation level under a constant fine system,which can be directly compared to the estimatesin Column 2. As required by theory, the sum ofthe monetary elasticities (for wage, price and fine)is zero.


Table 2Elasticities of firm output and industry open access equilibrium populations, for variable and constant fine structures

Elasticities

Constant fine: open access population (Be)Variable fine

Firm-level output (q*) Open access population (Be)

With respect to:0.796−0.060 0.765Wage (w)

1.069Price (p) −0.837 −0.9170.040 0.152Fine (r) −1.0090.244−1.024 0.347Detection (b)

Source: author’s calculations.

These models use the theory of the firm toutilise a considerable amount of primary data onthe economic conditions facing poachers. Al-though the results from the long-run analysis haveto be taken with some degree of caution, they dosuggest that the variables would have to changeconsiderably for any significant change in equi-librium elephant numbers to occur, despite theapparently large elasticities reported in Table 2.This is because these values are measured at the(very low) long-run equilibrium population level.Even if prices were to be halved, implying a40–50% increase in population, this would stillgive a very low population when compared to the1985 level. MGLW also estimate the implicationsof sole-ownership of the resource, as they suggestthat it is possible for there to be some potentialfor control and exclusion of effort. Using thelinear fine and linear production function in anintertemporal optimisation framework leads tosimilar conclusions as those for the open accessmodel at long-run equilibrium: the populationlevel of elephant is forced down to a very lowlevel, and there have to be considerable changes inthe ivory price before there are significant changesin that equilibrium level.

4. Aggregate simulation models

An alternative approach to the issue of poach-ing effort is to develop aggregate simulation mod-els of elephant population numbers. There have

been a number of models used (Pilgrim and West-ern, 1986; Milner-Gulland and Mace, 1990; Mil-ner-Gulland and Beddington, 1993) but here weconcentrate on an extension to the model pre-sented in Burton (1994b), which in turn is adevelopment of Caughley et al. (1990).

At the continent level, one of the few pieces ofdata that is available is an estimate of the totalharvest of ivory. Caughley et al. (1990) present amodel which consists of three equations, whichdetermine population levels (measured as tonnesof ‘live’ ivory), ivory harvest and hunting effort.The population dynamics are governed by a logis-tic equation (Eq. (11)):

Nt+1= (Nt−Yt) exp[rm(1−Nt/K)] (11)

where Yt is the harvest in year t, rm is the intrinsicrate of increase and K is the environmental carry-ing capacity. Harvest yield is a linear function ofeffort (E) and population level (N), with effortnormalised so that the ‘catchability coefficient’ isunity (Eq. (12)), i.e.

Yt=EtNt (12)

The level of effort is assumed to increase expo-nentially, at a rate f, from an initial level:

Et= (Y1950/N1950) exp( ft) (13)

where t is a time index with a value of 0 in 1950.The value of rm is taken to be equal to 0.06,(Hall-Martin, 1980; Calef, 1988) and the initialpopulation level N0 (i.e. in 1950) is assumed to beequal to K, the environmental carrying capacity.


The time series data on ivory production is derivedfrom Parker (1979, 1989). There is no consistenttime series data available on effort or populationlevels. In order to overcome this difficulty, keyparameters of the model are calibrated so that thehistorical simulation performance of the modelwith respect to the harvest level meets certaincriteria. However, the model of effort has nobehavioural content, and guarantees that extinc-tion will occur. Burton (1994b) proposes an alter-native representation of the factors governingeffort levels, and here we continue that develop-ment.

The effort Eq. (13) can be replaced by thefollowing effort equation (Eq. (14)), which is basedin the theory of open access exploitation of re-sources (see Hartwick and Olewiler (1986) for adiscussion in the context of fisheries).

Et+1=Et+gpt (14)

where pt is the profit per kilo of ivory harvested.Thus, if profits are positive, effort increases in thesector (at a rate governed by g, a parameter to beestimated), whereas negative profits cause it todecrease. Profit is defined as [IPtYt−ct Et ]/Yt,where IPt is the price of ivory, and ct the per unitcost of effort. Empirically the measurement of thisprofit variable is problematic, with little availabledata. However, Milner-Gulland and Leader-Williams (1992) give data that can be used toestimate the profit per kilo in Zambia for 1985,with an estimate of the price of 285 kwacha kg−1,and a cost of 22.3 kwacha kg−1 (equivalent to $50and $3.9, respectively). This information is used tocalibrate the level of profit simulated in the currentmodel. IP is defined as the real price of imports ofraw ivory into Japan (based on the data reportedby Barbier and Burgess (1989) and deflated by theJapanese consumer price index), normalised tohave a value of 285 in 1985. The real unit cost ofeffort, ct, is assumed to be a constant and becomesa parameter to be estimated. However, the cost perkilogram of ivory harvested is given by ct/Nt, andhence will change inversely as elephant numberschange. Within the calibration ct is constrained toequal 22.3*N1985, and hence the simulated cost perkilo generated by the model in 1985 is constrainedto equal 22.3 kwacha ($3.9), so that the level of

profit simulated by the model is consistent with the1985 Zambia data. The level of profit simulated bythe model in other periods will obviously vary,increasing as the real price of ivory rises, falling aselephant numbers fall and costs per kilo rise. Thismodel of profit will only represent a crude proxyfor the incentives that have motivated effort overthe period 1950–1986, but given the magnitude ofthe changes in the real price of ivory, it is probablya good enough indicator of general trends. Itshould be emphasised that the purpose here is toinvestigate the implications of an alternative spe-cification for the equation describing effort, ratherthan present a definitive model of it.

There are then two unknown parameters in thesystem: the initial population level, N1950, and g,the response of effort to profit. By utilising GAMS,a non-linear programming package (Brooke et al.,1988), it is possible to identify the values of theseparameters which minimise an objective functiondefined as (Eq. (15)):

%1987

t=1950

(Yst −Yt)2+b(Ns

1981−N1981)2 (15)

The first term is the sum of squared deviationsof the simulated yield from the actual yield over theperiod, and is the criteria used in Burton (1994b).But, as was noted there, the calibration can resultin simulated live ivory stocks that are considerablyin excess of field estimates. If time series data onpopulation levels were available they could beincluded, but generally this is not the case. How-ever IUCN/WWF/UNEP (1982) does provide anestimate of 1.2 million elephant for 1981, and thisinformation is used. Here the population level isconverted into live ivory by using an estimate of 1tonne of ivory per 100 elephant, and the coefficientb is chosen so that a 10% error in the simulatedpopulation in 1981 has the same impact on theobjective function as a 10% error in the yield in1981. On this basis, the single observation onpopulation levels is given the same weight as anysingle observation on yield. Given the high marginof error in the original population estimate and thepoor estimate of the translation from populationto live ivory, this is probably appropriate.1

1 The model was also solved with b increased by a factor of10, and this had relatively little impact on the results.


When calibrated on this basis the estimate ofN1950 is 29754 tonnes, and g=1.0693×10−5. Theimplicit value of c is 3.1×105. The simulatedvalue of the population level (of live ivory) in1981 is 16391 tonnes, above the point estimate of12000 tonnes but well within the margin of error.The increase in effort between 1950 and 1987 isequivalent to a 5.9% growth rate, not dissimilar tothe 5.73% growth rate estimated by Caughley etal. (1990), but obviously the long-run behaviourimplied by the two systems is quite dissimilar.Here, as population levels fall then the cost perunit of ivory harvested rises, profits fall and effortleaves the industry. At 1987 ivory prices the equi-librium population level is 1121 tonnes of ivory(equivalent to approximately 112000 elephant).Simulation of the model post-1987, holding theprice of ivory at its 1987 value, indicates that thesystem converges on that equilibrium but thedynamics are such that the population falls tovery low levels in the interim. One cause of this isthe relatively slow response of effort to profit, sothat there is a very gradual exit of effort as profitsbecome negative, and hence populations continueto decline. This may be purely an artifact of themodel, however, as it has been calibrated entirelyover a period when profits are positive: the situa-tion of negative profits is completely outside ofthe historical experience.

Fig. 2 gives the estimated phase diagram for themodel. The vertical line indicates the locus ofpoints where economic equilibrium is achieved

(i.e. the change in effort is zero), while the secondline indicates the locus for biological equilibrium(i.e. where there is no change in population lev-els). The vectors in each quadrant indicate thedirection of change of the system when out ofequilibrium, and reveal that the equilibrium pointis stable.

A further extension of the model is to considerthe change in technology over the period. AsDouglas-Hamilton (1987) notes, there has been asubstantial increase in the availability of modernfirearms in the range states, and Milner-Gullandand Leader-Williams (1992, p. 399) suggest thatthese increase the number of kills of a gang ofpoachers by a factor of 16. We accommodate thispossibility by introducing a time varying catcha-bility factor, so that the yield equation becomes:

Yt= (0.1+0.9 exp(−dt))−1EtNt

The time trend t takes a value of zero in 1950,so the catchability coefficient is equal to 1 in 1950,and then increases up to a limit of 10. This is aconservative upper limit, and it allows for anincomplete diffusion of the new technology by1987. The change in catchability coefficient isgiven by the coefficient d, which is constrained tobe positive. When the model is calibrated usingthis yield equation the initial population is esti-mated at 27168, g=2.599×10−6, with an im-plicit value of c of 1.046×106. d is 0.055, whichresults in the catchability coefficient increasing to4.46 by 1987. The level of effort also increases,but at a much lower rate than before (equivalentto 2.66% per year: see Fig. 3 ). This distinctionbetween changes in effort and advances in tech-nology may not appear to be a significant one,especially as the simulated time path of yields isquite similar (Fig. 4), but it does have implica-tions for the long-run equilibrium population. Ifone holds both the ivory price and the technologyat 1987 values then the equilibrium value is 848tonnes of live ivory, some 25% lower than theprevious estimate. If one allows the catchabilitycoefficient to rise to a value of 10, then it falls to378 tonnes. What is also notable is that the esti-mated change in effort over the period is alteredby the representation of technology, and the re-sponse of effort to the presence of profit is re-Fig. 2. Phase diagram for open access simulation model.


Fig. 3. Simulated effort levels. Note: E1 is derived from asimulation where hunting technology is assumed to be con-stant over the period, while E2 is derived from a simulationwhere hunting technology is allowed to advance. See text fordetails. For the purposes of this figure, both indices of effortare normalised to equal one in 1950.

pose of the ban on trade was to reduce incentivesto poach by restricting final demand. The price ofivory has to fall by some 90% if the simulated1987 population is to be maintained in the con-stant and variable technology models (if the latteris held at 1987 technology levels: an even largerdecline is needed in the variable technology modelif the catchability coefficient is increased to avalue of 10). This is not very encouraging, in thatit implies that the effectiveness of the ban at thelevel of final demand has to be high if it is to havethe required effect on the poachers.

5. Discussion

Any analysis of the economics of ivory poach-ing is going to be constrained by the paucity ofgood data available, and in particular, time seriesdata. This paper has looked at various ap-proaches to the question of the relationship be-tween poaching and economic incentives. Withfew direct measures of poaching effort one isreduced to indirect measurement of the links be-tween economic incentives, enforcement activityand levels of poaching, and then inferring theresponsiveness of the underlying poaching effort.

Those studies that look directly at the determi-nants of poaching activity (i.e. those in Section 2)all conclude that there is little evidence of anyimpact of the ivory ban in 1989, given the datathey have to work with. The simulation models(Sections 3 and 4), by construction, contain apositive relationship between poaching effort andprices, and hence one has a primea facia case forthe ban having a positive effect on elephant con-servation. However, what those models do revealis that significant poaching effort will remain evenif the ivory price is reduced substantially, andhence suggest that if the ban is to be effective, thelink between final consumer and poacher has tobe practically absolute: any slippage may be suffi-cient to give sufficient incentives to poachers toprevent recovery in elephant populations. In thiscontext, the comment reported in Dublin et al.(1995) that ‘‘..donor countries in the north seemto feel that the problem of elephant poaching hasbeen ‘taken care of’ by the ban and that there is

duced by some 75% as a result of using thevariable technology model (Fig. 4).

These simulations take no account of the po-tential impact of reduced supply of ivory onprices. Barbier and Burgess (1989) provide esti-mates of elasticities of demand for ivory whichcould be used to generate flexibilities, but Milner-Gulland (1993) has raised doubts about these(and suggests that the elasticity of demand may bezero). Furthermore the simulated harvests nearequilibrium are so small they may not provideuseful information on price responses. Of perhapsmore relevance is the ivory price that is requiredto maintain certain population levels, as the pur-

Fig. 4. Actual and simulated ivory yields. Note: Y1 is derivedfrom a simulation where hunting technology is assumed to beconstant over the period, while Y2 is derived from a simula-tion where hunting technology is allowed to advance. See textfor details.


Table 3Prices for ivory from hunter to first middleman US$/kg

Pre-ban Post-ban

Cameroon4Savanna 2

34–62Forest 86

Tanzania 10–266–12Malawi 21–4252–87

36–4046–56Ivory CoastZaire 16 6

21–4252–87Malawi18–24Zambia 14–26

Source: Dublin and Jachmann (1992).

Although the structure of the ivory market maybe known in general terms there is relatively littlework on quantifying the structure, and this infor-mation is needed if the economic activity ofpoaching is to be controlled. The determinants ofpoaching effort is perhaps the most developedarea, but as we have seen, the limitation of poordata places extreme restrictions on what can beachieved. Without that information, identifyingthe effects of the trade ban (both ‘ex ante’ and ‘expost’) is fraught with difficulties and runs the riskof ‘‘..equating association and correlation withcausality, failing to identify and cut through con-founding factors, failing to replicate, failing tobalance, failing to control’’ (Caughley, op cit p.239). Given that the African elephant listing isoften viewed as one of the more successful exam-ples of the CITES listing mechanism, the difficultyin showing its effect is of concern. Without estab-lishing this link the objective development of pol-icy is severely hampered, and evaluating evenincremental management, such as the recentCITES downlisting of local elephant populations,and the associated requirement to assess any neg-ative impacts of the resumption in trade on otherrange states will be problematic. The adoption bythe parties in 1997 of a Resolution to establish ‘‘acomprehensive, international monitoring system...with the objective of measuring and recordingcurrent levels and trends of poaching and illegaltrade in ivory in African and Asian range states,and in trade entrepots’’ and assessment as to‘‘whether, and to what extent, observed trends arethe result of changes in the listing of elephantpopulations in the CITES Appendices and/or theresumption of legal international trade in ivory’’(Conf.10.10, cited in Gray, 1997), may go someway to filling this gap, but without an understand-ing of the historical patterns of exploitation (andthe associated data) such an analysis will beincomplete.

Given that the management of elephants, andthe ivory trade ban, have probably received moreattention from the social scientists than any otherspecies, the implications of this paper for themanagement of other species which are underthreat are not hopeful.

no need to continue supporting law-enforcementefforts in the field’’ is particularly worrying. Con-sistent data on prices is difficult to come by, butthose reported by Dublin and Jachmann (1992)(summarised in Table 3) suggests that prices havenot fallen everywhere, implying a residual demandfor ivory, either held speculatively within Africaor smuggled illegally out of the continent. Evenwhere prices have fallen, the results outlinedabove would suggest that changes in prices of thismagnitude are not sufficient to significantly re-duce the level of poaching effort.

6. Conclusions

Apart from the specific inferences drawn fromthe different models reviewed, one further com-mon theme from them all is the paucity of appro-priate data with which to undertake the analysis.Caughley (1994), in a review of conservation biol-ogy, outlines the steps taken to protect the LordHowe woodhen, found only on the South pacificisland of the same name. He suggests that thesteps taken—diagnose the agent of decline, neu-tralise the agent of decline, re-establish the speciesof concern—may serve for almost any other trou-bled species. In the case of the African elephant,diagnosing the agent of decline is relativelyclearcut, but neutralising the agent of decline isnot as straightforward as shooting out a feral pigpopulation, and requires an understanding andmanipulation of the economic incentives facingthe poacher.


Acknowledgements

Part of the research reported here was sup-ported by an Area Studies Grant, University ofManchester, UK, and a grant from The NuffieldFoundation. The reviewers of an earlier draftmade helpful comments on exposition andcontent.

References

Barbier, E.B., Burgess, J.C., 1989. Analysis of the demand forraw ivory: case studies of Japan and Hong Kong. LEECpaper 89-05 IIED/UCL London Environmental EconomicCentre, London.

Barbier, E.B., Burgess, J.C., Swanson, T.M., Pearce, D.W.,1990. Elephants Economics and Ivory. Earthscan, London.

Barnes, J.I., 1996. Changes in the economic use value ofelephant in Botswana: the effects of international tradeprohibition. Ecol. Econ. 19, 215–230.

Bonner, R., 1993. At the Hand of Man. Simon and Schuster,London.

Brooke, A., Kendrick, D., Meeraus, A., 1988. GAMS: AUsers Guide. The Scientific Press, Redwood, California.

Bulte, E.H., van Kooten, C.G., 1996. A note on ivory tradeand elephant conservation. Environ. Dev. Econ. 1, 433–443.

Burton, M.P., 1994a. Conservation expenditure and conserva-tion success. Working Paper in Agricultural EconomicsWP94/01, University of Manchester, Manchester.

Burton, M.P., 1994b. Alternative projections of decline of theAfrican elephant. Biol. Conserv. 70, 183–188.

Calef, G.W., 1988. Maximum rate of increase of the Africanelephant. Afr. J. Ecol. 26, 323–327.

Caughley, G., 1994. Directions in conservation biology. J.Anim. Ecol. 63, 215–244.

Caughley, G., Dublin, H., Parker, I., 1990. Projected declineof the African elephant. Biol. Conserv. 54, 157–164.

Douglas-Hamilton, I., Michelmore, F., Inamdar, A., 1992.African Elephant Database. United Nations EnvironmentProgramme.

Douglas-Hamilton, I., 1987. African elephants: populationtrends and their causes. Oryx 21, 11–24.

Dublin, H.T., Jachmann, H., 1992. The impact of the ivorytrade ban on illegal hunting of elephants in six range statesin Africa. WWF International Research Report, GlandSwitzerland.

Dublin, H.T., Milliken, T., Barnes, R.F.W., 1995. Four yearsafter the CITES ban: Illegal killing of elephants, ivorytrade and stockpiles. A report of the IUCN/SSC AfricanElephant Specialist Group.

Gray, J., 1997. Report of the tenth meeting of the conferenceof the parties to CITES. In Traffic Bulletin 17, 5–19.

Hall-Martin, A., 1980. Elephant survivors. Oryx 15, 355–362.Hartwick, J.M., Olewiler, N.D., 1986. The Economics of

Natural Resource Use. Harper and Row, New York.IUCN/WWF/UNEP, 1982. Elephants and Rhino in Africa—a

Time for Decision. IUCN, Gland.Ivory Trade Review Group, 1989a. The Ivory Trade and the

Future of the African Elephant. Volume 1: Summary andconclusions.

Ivory Trade Review Group, 1989b. The Ivory Trade and theFuture of the African Elephant. Volume 2: Technicalreports.

Jachmann, H., Billiouw, M., 1997. Elephant poaching and lawenforcement in the central Luangwa Valley, Zambia. J.Appl. Ecol. 34, 233–244.

Khanna, J., Harford, J., 1996. The ivory trade ban: is iteffective? Ecol. Econ. 19, 147–155.

Kreuter, U.P., Simmons, R., 1995. Who owns the elephants?The political economy of saving the African elephant. In:Anderson, T., Hill, P. (Eds.), Wildlife in the Marketplace,Chapter 8. Rowman and Littlefield, Lanham MD.

Leader-Williams, N., 1993. The cost of conserving elephants.Pachyderm 17, 30–34.

Milner-Gulland, E.J., 1993. An econometric analysis of con-sumer demand for ivory and rhino horn. Environ. Resour.Econ. 3, 73–95.

Milner-Gulland, E.J., Leader-Williams, N., 1992. A model ofincentives for the illegal exploitation of black rhinos andelephants: poaching pays in Luangwa Valley, Zambia. J.Appl. Ecol. 29, 388–401.

Milner-Gulland, E.J., Beddington, J.R., 1993. The exploitationof elephants for the ivory trade: an historical perspective.Proc. R. Soc. Lond. B 252, 29–37.

Milner-Gulland, E.J., Mace, R., 1990. The impact of the ivorytrade on the African Elephant Loxodonta africana popula-tion as assessed by data from the trade. Biol. Conserv. 55,215–229.

Parker, I.S.C., 1979. The ivory trade. Report to US Fish andWildlife Service. Washington.

Parker, I.S.C., 1989. The raw ivory trade 1979–1989. Reportto CITES, Lausanne.

Parker, I., Graham, A.D., 1989a. Men, elephants and competi-tion. Sym. Zool. Soc. Lond. 61, 241–252.

Parker, I., Graham, A.D., 1989b. Elephant decline: downwardtrends in African elephant distribution and numbers. PartI. Int. J. Environ. Stud. 34, 287–305.

Parker, I., Graham, A.D., 1989c. Elephant decline: downwardtrends in African elephant distribution and numbers PartII. Int. J. Environ. Stud. 35, 13–26.

Pilgrim, T., Western, D., 1986. Inferring hunting patterns onAfrican elephants from tusks in the international ivorytrade. J. Appl. Ecol. 23, 503–514.

Sugg, I., Kreuter, U., 1994. Elephants and Ivory: Lessons fromthe Trade Ban. Institute of Economic Affairs, London.

.

An assessment of alternative methods of estimating the effect of the ivory trade ban on poaching...

Documents

Transcript of An assessment of alternative methods of estimating the effect of the ivory trade ban on poaching...