The Flower of Paradise: Substitution or Income Effect? Sara Borelli University of Illinois at Chicago

Transcription

1 The Flower of Paradse: Substtuton or Income Effect? Sara Borell Unversty of Illnos at Chcago Abstract The am of ths paper s to evaluate the mpact of a wage ncrease on Qat consumpton for the Djboutan mal household head usng a statc framework. In partcular, by usng a system of smultaneous of equatons wth Qat consumpton and Wage as endogenous varables, the paper estmates whether a wage ncrease wll affect Qat consumpton. Our fndngs suggest that the mpact of wage on Qat consumpton s postve and sgnfcant. Gven the realstc assumpton that the Qat s a tme ntensve commodty, ths means that ncome effect (generate by a wage ncrease) wll preval on the substtuton effect. *JEL Classfcaton: I12, D12 *Keywords: Qat, Djbout, Tme Intensve Good. Ph.D. student Unversty of Illnos at Chcago, 601 S. Morgan Street, Chcago Emal: sborel2@uc.edu Telephone/Fax: / I would lke to thank Davde Furcer for the useful comments. I, alone, am responsble for any errors. 1

2 Emerald mage t sweetens my heart makng my state of mnd and my days so pleasant. Qat s everythng that one could wsh for; t brngs good and drves evl away. The good-n-heart have wrtten the name of Allah n ts leaves. Qat unvels the mystery of Allah. (Yemen poem XVI century). 1. Introducton Some of the frst analyss of addctve behavour can be found n Becker and Stgler (1977) and Becker and Murphy (1988), who systematcally develop a model of addctve behavour n an nter-temporal framework. They defne addcton n a broad sense: people get addcted not only to alcohol, cocane and cgarettes but also to work, eatng, musc, televson, ther standard of lvng, other people, relgon and many other actvtes. The man results of the paper show how addctve consumpton affects ntertemporal behavour and clam that addctons, even strong ones, are usually ratonal n the sense of nvolvng forward-lookng maxmzaton wth stable preferences. The purpose of the followng paper s to evaluate the mpact of drug consumpton on the allocaton of tme usng a statc analyss. Drug use does not affect only the way n whch people trade-off present and future consumpton, but t affects also the way people respond to changes n the value of tme (.e. wages) n a statc framework. The analyss s carred out by means of the theoretcal model developed by Becker n an earler paper (Becker, 1965) and by usng 1996 data for the Afrcan socety of Djbout. In the Horn of Afrca the consumpton of Qat (a substance smlar to amphetamne) s an expensve habt of the male 2

3 populaton. Qat s perceved as a desrable good by the consumers but t s also a prvate bad because harmful to the personal health. By usng the (realstc) assumpton that Qat s a tme ntensve commodty, we want to nvestgate f there s a postve or negatve dependence of drug use on wage. The paper s organzed as follows. Secton 2 descrbes the consumpton of Qat n Djbout, secton 3 presents the theoretcal model, secton 4 descrbes the data, secton 5 presents the econometrc model and the results. Secton 6 reports the man conclusons. 2. Qat n Djbout The name Qat comes from the plant Catha Eduls Forskal, a plant growng at meters of alttude. Mountans of Ethopa and South Araba ft partcularly well the clmatc condtons necessary for ths plant to grow and survve. To keep ntact ther propertes, Qat leaves have to be consumed n maxmum three days after they are taken away from the crops. Qat consumpton n ths regon has a long hstory. The frst avalable wrtten reference dates back to In ths perod ts stmulant effects are well known, and t s beleved to be a plant of unordnary propertes. The ncreased consumpton of Qat n the last century has been subject of concern, and many studes have been undertaken by the WHO to senstze medcal envronments. To characterze the strong socal and cultural mpact of ths knd of consumpton n Djbout, socologsts often use the word "socal plague". Qat consumpton s prevalently a male habt that heavy affects household budget. The lfe of a Qat chewer plays around the socalled Qat partes held n the afternoon untl dusk. Qat sales start n the early afternoon 3

4 when an arplane comng from Ethopa arrves brngng daly 12 tons of Qat. Streets become empty for many hours durng afternoon and the cty falls asleep. Thus, Qat consumpton s a really wdespread and pervasve phenomenon. Moreover, the commerce of Qat s one of the most consstent sources of fscal revenue for the Government (t consttutes on average between 9% and 15% of mport taxes). The proporton of these taxes s very hgh f we consder that the Djboutan economy s very poor n tradable natural resources. The hgh proporton of resources allocated to Qat s at the expenses of basc household expendtures lke food and educaton. In the followng, we dscuss the queston of addcton related to Qat consumpton. In medcal terms, the noton s that true addcton s sgnalled by the experence of physcal wthdrawal symptoms when the drug s wthheld from the habtual user. A concomtant dea s tolerance, the noton meanng that as the body becomes accustomed, a contnuously ncreasng need s felt for more of a substance, wth consequent drven seekng to procure the needed fxes. Then, ths leads to contnually greater neglect of ordnary behavours n favour of drug experence and further drug seekng. In 1956 a group of experts of the WHO defned seven categores of drug dependence, and among them Qat type s a self-defnng category. It contans Cathnone, a substance that seems to have a hgh stmulant capacty and produces effects smlar to those of amphetamnes (though t s structurally dfferent); t ncreases heart rate, locomotor actvty and oxygen consumpton. Some studes report wthdrawal symptom and tolerance assocated wth Qat consumpton (Nancn et al.), but, as reported by Wer (1980), a serous defect of these studes s that there s no menton of any montorng of a control group of non-chewers n dentcal condtons. In an mportant major research project on Qat consumpton was carred 4

5 out n Yemen under the drecton of John Kennedy. The possble stmulant and addctve propertes of Qat were nvestgated to assess the major effects of ths nsttutonalzed drug upon the socal lfe, economcs and health of Yemen people (Kennedy et al. 1980). The results show that Qat does not produce wthdrawal symptom, even f the author ponts out that a mld form of physologcal dependence does result from extremely heavy use. Wth regard to the queston of tolerance, the results do not seem to be uncontroversal. Other authors (such as Wer, 1980) stress the mportance of analyzng Qat consumpton n a broader perspectve. Even f the medcal queston of addcton does not seem to be completely uncontroversal, t s wdely recognzed that socal and economc factors are of prmary mportance n determnng ndvdual consumpton levels, and account for ndvdual dfferences n the consumpton of Qat: f the pleasurable and addctve propertes of a drug are known to be powerful, consderable weght can be placed on medcal, physcal explanatons for ts consumpton. But n the case of weak substance such as Qat, we must lean more heavly on socal explanatons for ts popularty and expense Obvously Yemen chew Qat socally because of ts stmulant effects but hgh expendture on Qat and ts ncreasng popularty are better understood f they are seen more as a reflecton of the value Yemen place on the socal effects of beng Qat consumer (Wer, 1980). These aspects strongly characterze the consumpton of Qat also n Djbout, and they wll be used to support the assumptons stated n the theoretcal model. 5

6 2. The Theoretcal Model A wde range of medcal, socologcal and psychologcal lterature deals wth the problems of addctve 1 consumpton, manly alcohol and drugs. But why do people consume some knd of goods? Maybe, t could seem napproprate to analyse ths problems from an economc pont of vew, but we beleve that the nsghts provded by other dscplnes can gve useful suggestons to understand the economcs of addcton n general sense, and ts relevant mplcatons for observed economc behavours. Psychologsts, n partcular, descrbe a wde range of reasons nducng people to develop an addctve behavour despte ts negatve health effects. A common element of analyss s based essentally on the dea that drug and alcohol consumpton can gve pleasure, provde help n facng daly problems, easy the communcaton and the nteracton among people and wthn the communty n whch one lves (Salvn et al. 1998, 2002). The medcal term toxcophla, n partcular, ndcates a class of behavours organzed and ntentonally fnalzed to reach the mental and physcal status nduced by the use and abuse of substances. As descrbed n the prevous sesson, the socal aspect of drug consumpton s very strong n the case of Qat. As reported by Kennedy (1987) the meanng of chewng Qat s that t gves the chewer a meanng, a sense of hs or her own dentty n rapdly changng world over whch he or she has lttle or no control The modern Yemen has a need to dentfy hmself n a postve way. 1 Note that her I am usng the term addcton how t s perceved n the common sense wthout any specfc reference to medcal characterzaton that wll be ntroduce later n the paper. 6

7 Thus, n a very smple way, we could say that Qat consumers feel what we defne a utlty gap ( γ ) dervng from the need to be part of a socety where ths knd of consumpton represents a very mportant mean of socal ntegraton. Thus, the utlty functon of an ndvdual can be represented n the followng way: U = U ( Z... Z ;γ ) 1 N where Z, = 1... N s a consumpton good wth U > 0, U < 0. Note that γ s not Z Z Z drectly observable from a economc pont of vew, but when ths utlty gap reaches some ndvdual threshold level, we could observe some changes n the consumpton pattern. Through Qat consumpton (and n general other drugs) the ndvdual replaces the utlty gap wth a new D good that drectly enter the utlty functon and provdes subjectve satsfacton through a drug experence (ths s what s called euphora n Stgler and Becker s terms (1977)). In fact, the frst expermental effects of Qat are a gradually developng mld euphora, alertness, and feelng of contentment, confdences, and gregarousness 2. Thus, by consumng Qat people attempts to reach a state of general well beng. However, the euphorc state can be followed by a state of confuson, mspercepton of tme and space and hallucnatons. 2 Qat consumers seem to consder t as a real good because of ts stmulatng propertes, and do not take nto account the negatve effects on ther health. The habt of chewng Qat causes nsomna, anorexa, and constpaton. Qat consumers are more susceptble to the contracton of venereal dseases, heart dseases, tuberculoss and others. Qat consumpton s also assocated wth hgh tobacco consumpton: durng ther meetng, Qat users usually smoke 1-2 packages of cgarettes. 7

8 We assume that the Z goods and the drug experence good D enter the utlty functon through a productve process that uses as nputs tme and market goods. As Becker ponts out a more ntutve assumpton s that tme and goods enter nto the producton of commodtes whch drectly provde utlty. These commodtes cannot be purchased n the marketplace but are produced as well as consumed by households usng market purchases, own tme and varous envronmental nputs. These commodtes nclude chldren, prestge and esteem, health, altrusm, envy and pleasures of the sense (Becker, 1981). In the present framework we gnore the consumpton captal aspects of drug consumpton whch allows establshng a relatonshp between past and current consumpton. Ths s an mportant lmtaton of the followng analyss that can be n part justfed by the data avalable that are not adequate to estmate a dynamc model, and n part by the nature of the drug experence assocated wth Qat. As descrbed n prevous sesson, the socal dmenson of addcton s very strong n the case of Qat, and the socal nterpretaton of the assocated ambguous physcal experence makes t possble to derve pleasure from ths actvty. The ndvdual optmsaton process can be descrbed as follows: ( Z Z D) max U = U..., 1 N (1) subject to Z = z( x, T D = d( Q, T Z D ) ) (2) n T = TZ + TD + Tw = Tc + T =1 w (3) 8

9 p x x + pqq = m + wtw (4) where equaton (3) represents the tme constrant descrbng how the total tme s dvded between consumpton tme,.e. the tme spent n producng the goods drectly enterng the utlty functon ( N =1 T + T = T ), and tme spent at work. The frst constrant descrbes how Z D c the nput vector of market goods x (=1,,N), Qat (Q) and the vector of tme nputs (j=,d wth =1,,N) are used n the producton of the Z and D goods. The last one s the usual budget constrant that dstngushes between labor and non-labor ncome (m). We assume producton functons homogenous of frst degree wth T j Z T Z Z D D > 0, > 0, > 0, > 0, x T T D D and specfyt D = tq, where t s the amount of tme necessary to chew one unt of Qat. We can combne the tme, budget and technologcal constrant n a more compact form: ( px x + wtz ) + ( pq + wt) Q = wt + m. (5) The LHS of equaton (5) represents the full expendture and π = wt s the full prce D p q + of the D good. Thus, the full prce s composed by two dstnct elements, the out-of pocket 9

10 (or drect) prce p q for the nput bought on the market, and the tme or ndrect prce wt, whch expresses the cost of tme n terms of foregone earnngs nvolved n the producton of the drug experence. Thus, the full ncome (RHS of equaton 2) s gven by the sum of the money ncome achevable f all the tme avalable were devoted to work and the nonlabor ncome. The total, or full, budget constrant descrbes how the full ncome s spent on the commodtes ether drectly, through the expendtures on market goods, or ndrectly through the forgong of ncome,.e. by usng tme at consumpton rather than at work. From the frst order condtons of the consumer s optmzaton problem we get 3 : U Z = λπ Z =1,,N (7) U D = λπ, (8) D whch mples: MRS DZ = π π D Z = p p q x dq tdq + w dd dd dx dtz + w dz dz. (9) 3 Maxmzng the utlty functon, subject to the full ncome constrant, usng the fact that the producton functons are homogeneous of the frst degree gves: U Z U D dx dtz = λ p x + w = λπ dz dz dq dtd dq tdq = λ pq + w λ pq w = λπ D dd dd = + dd dd, Z where s the Lagrange multpler and the full prces π Z, π D are determned by the market good and tme nputs requrement necessary to the producton of one unt of the Z and D goods. 10

11 Thus, usng ths approach, we can dstngush the dfferent commodtes based on ther ntrnsc degree of tme and good ntensty 4 to nvestgate the effect of an ncrease n the wage. As descrbed n the prevous sesson, Qat consumers spend a large fracton of ther tme n chewng Qat wth frends. Men meet usually n the afternoon at the so-called Qat partes to spend together many hours untl dusk. Thus, we can reasonably assume that the producton of the typcal Djboutan drug experence s a relatvely tme ntensve actvty. Ths can be formalzed as follows: Assumpton 1. Tme ntensty: w p q dtd dq > w p x dt dx x Ths means that f we look at two hypothetcal soquants of producton for the D and Z good, we have more tme (n real terms) emboded n the producton of the drug experence than n the producton of other goods. Note that f we assume that the drug experence provdes drect satsfacton, we cannot drectly measure the amount of euphora produced. What we can observe s the use of the (nput) Qat. However, gven the above specfed producton functon of D, we can determne a one to one correspondence between the use of Qat n the producton of D and the good D. Ths s equvalent to say that the Qat users dentfy and drectly assocate the output of the productve process (D) wth the nput, so that by statng that the producton of euphora s a tme ntensve actvty s equvalent to 4 t Becker (1965) defnes the relatve margnal mportance of tme as π. 11

12 say that Qat consumpton s a tme ntensve actvty. In psychologcal-techncal terms, ths could be vewed from a behavoural perspectve, whch consders the repeated use of substances as behavour determned by the percepton of the physologcal alteratons produced by the substance tself. Moreover, the fact that the person assocates a postve meanng to these perceptons has a functon of renforcement on the use or abuse (Turch, 2002). Gven the role Qat plays as socal aggregator n the lfe of the typcal male Djboutan communty, we could reasonably say that ths knd of consumpton generates a cultural or socal form of dependence. Assumpton 2. Addcton: Qat does not generate addcton n the strct medcal sense but generates a form of socal dependence. Gven the nature of ths knd of consumpton t s therefore nterestng to nvestgate the effects of ths nsttutonalsed stmulant on the schedulng and content of socal actvtes, the structure of economc lfe and even the percepton of realty. The frst step n ths paper s to analyse the consumpton of Qat n the theoretcal framework developed above. Gven the assumpton of tme ntensty, we want to analyse 12

13 the effect of an ncrease n the value of tme, represented by an ncrease n wages. From equaton (9) we have 5 : w p q dtd dq > w p x dt dx x π D π Z w > 0. (10) Thus, an ncrease n wages determnes a relatvely large ncrease n the full prce of the relatvely more ntensve commodty D. The mportance of consderng the full prce becomes here evdent, because t allows dstngushng goods n terms of the amount of tme necessary to ther producton. An ncrease n the wage wll determne therefore two effects. On one sde, by the substtuton effect, an ncrease n the value of tme wll determne a shft away from the producton of D (or equvalently from the consumpton of Qat) tme ntensve commodty to the good ntensve ones, snce now the opportunty cost of spendng tme at Qat partes s hgher. On the other sde we have an ncome effect, whch operates 5 Takng the dervatve of (9) wth respect to w we get: dq tdq dtd dx dq dt p + q w p x pq π D = dd dd dd dz dd dz = w 2 π Z w dx dtz p x + w dx dtz p x + w dz dz dz dz Ths mples that : π D dx dtz sgn = sgn p. xt pq w π Z dz dz Assumpton 1 mples that ths term s postve. In fact we have that dtz dtd dtz dx dtz t dx dq dx dtd p xt pq >0 > > snce t =. dz dz pq p x pq p dq x Z = p x tdq dd p x dx dz dx dz p q dq dd dt + w dz Z dt dz 2 Z 13

14 exactly n the opposte drecton. The former would reduce Qat use, the second would ncrease t, and the total effect wll depend on the relatve strength of the two. 4. Data The cross-secton data set s the natonal household survey (EDAM, Enquête auprès des ménages) conducted n 1996 by the Statstcs Department of Djbout (DINAS) among the sedentary populaton. Four man groups represent the populaton: nomads, homeless, refugees and sedentary trbes, that s the group targeted by the EDAM. The survey was artculated n separate questonnares for households and ndvduals. The survey sampled 15,701 ndvduals belongng to 2,380 famles. About 65 percent of the households lve n the cty captal Djbout, 16 percent n other towns and the remanng 17 percent n rural areas. Data on Qat use reports weekly consumpton and the number of users n the household. Because Qat s perceved as a good rather than a bad, the consumpton s revealed truthfully. For the purpose of the paper, we selected a subgroup of 1,787 maleheaded households. Male household head are consdered the representatve consumers of Qat. Qat s n fact a male habt, whle female thnk that husbands who uses Qat are unbearable and clam that they spend ther earnngs on the prvate consumpton of Qat rather than satsfyng the famly needs. A 1984 survey of 498 households n Djbout cty (Republque de Djbout, 1984) reported that 75 per cent of household heads consumed Qat every day. The selected subgroup s composed by Qat users and non Qat users, respectvely the 56 and 44 percent of the sub-sample (Table 1). Inspecton of Table 2 reveals that on average the largest household expendture tem s food, takng 45 percent of the budget. Qat 14

15 represents 10 percent of the total household expendture, energy s 7 percent, educaton 2 percent, clothng 3 percent and health 2 percent. The comparson of the average budget between Qateur and non-qateur Djboutan households reveals that the Qat share reaches a level of 20 percent n Qateur households. 5. Econometrc Model and Results The purpose of our paper s to evaluate whether an ncrease n the value of tme reduces or ncrease the use of Qat. Thus, we want to estmate our drug functon to evaluate the effect of male household head s wage on drug use. The frst problem that arses n ths context s that drug use and wages are smultaneously determned. As we want to see how wages affect drug consumpton, t could be well the case that causalty runs n both drectons,.e. also drug use affect wages. The detrmental physcal and psychologcal effects of drug use suggest that wages should be affected by drug use. Another reason why a sngle equaton model s napproprate s due to the fact that there are unobservable characterstcs that affect drug consumpton and that could also affect wages. In the ntertemporal framework developed by Becker and Murphy, drug users are characterzed as more mpatent ndvduals,.e. they have a hgher rate of tme preference. But ths could also affect ther earnng profle, snce mpatent ndvduals are expected to choose a flatter experence earnng profle (Mncer, 1974). In the case of Qat, the (unobservable) socal pressure leadng to ts mass consumpton could affect the ndvdual decson to work and also the ndvdual choce of a partcular earnng profle compatble wth such tme ntensve actvty. 15

16 For these reasons we estmate a smultaneous system of equatons wth wages and drug use beng jontly determned. The model that wll be estmated n ths paper s: Q = α + α X 0 Q W = β + β X w + α W + u 2 d + β Q + u 2 w (11) where X Q s a set of explanatory varables n the Qat-drug equaton. Qat use s consdered a representatve consumpton of the male household but t s regstered at household level. In some sense, ths consumpton can be consdered as expresson of the power of the household head wthn the household and therefore of hs ablty to dvert resources toward the consumpton of an exclusve good. Thus, we nclude both ndvdual and household characterstcs n the drug equaton: household head s age, dummy f household head can read and make calculatons, years of formal schoolng completed, number of chldren 0-5 years old, number of chldren 6-18 years old, martal status, geographc dummes and weakly average household non-labor ncome (ths ncludes, among others, fnancal revenues, pensons, subsdes and n partcular transfers from frends and relatves abroad or n the country). The set of ndependent varables n the wage equaton (X w ) ncludes household head s age, age square, dummy f household head can read and make calculatons, years of schoolng, dummy varables for locaton and sector of employment. To estmate the model we use a Two-Stage Least Squares (2SLS) procedure. As measure of drug consumpton we use the (logarthm of) weekly Qat consumpton n grams. Qat s bought on the market n bundles of dfferent szes (for example 100, 200 grams). The regular consumer buys daly 300 grams of Qat. The occasonal consumer chews sx

17 grams bundles each day of the weekend. The average weekly Qat consumpton for the subsample of household males s 948 grams (or 1757 f only Qat consumers are consdered).we frst estmate a reduce form for the wage equaton usng the two step Heckman procedure (1976,1979) to take nto account that wages are observed only for workng household heads. The dependent varable use s the (logarthm of) weakly wage. Ths procedure allows gettng consstent estmates and can be used as a part of the 2SLS procedure. As a benchmark, we estmate also a reduced form for the wage equaton wthout the selectvty correcton. To estmate the Qat consumpton equaton we use both standard Ordnary Least Squares (OLS) and a Tobt regresson to take nto account that a fracton of the household heads does not consume Qat. To address the dentfcaton ssue of the system of structural equatons we have to assume that there are some varables affectng wages that do not enter the drug use equaton. The excluson restrcton on the drug functon s that the dummes for the sector of employment (transportaton and telecommuncaton, agrculture, fshery, servces and ndustry sector) and household head age square have no drect effect on Qat use. We tested for over-dentfcaton restrctons and we found that these varables can be excluded from the drug use equaton. The results of the estmates are reported n Table 3. They show that, n general, drug use s related to the age of the household head, ts educaton level, the sze of the household and hs wage. The sze of the household s an mportant factor: males n larger households spend more on Qat. When the household head s alphabetsed he spends more on Qat. The fact that he can read and make calculatons s not necessary related to ts level of educaton, and thus t does not mean that he s aware of the negatve effects of Qat consumpton but rather he could have some addtonal means to provde Qat for hmself. Hgher levels of educaton mply, 17

18 nstead, that the household head consume sgnfcantly less Qat. For our purpose, regardless of the method used, the estmates show a postve and sgnfcant effect of wages on Qat consumpton. Moreover, the margnal effect for the wage varable s approxmately the same for all the econometrc specfcatons (around 2). In terms of our model we can nterpret ths as evdence of the prevalence of the ncome effect over the substtuton effect. Gven the strong socal and cultural dependence of most male household heads from ths knd of consumpton, an ncrease n the value of tme s consdered as an mmedate way of havng more fun nstead than workng more to take advantage of the ncrease n wage. Any knd of addcton, even socal ones, s lkely to make ndvduals more mpatence, and ths s gong to affect not only ther nter-temporal behavour but also the way n whch they make ther choces at any gven pont n tme. Gven the assumpton that Qat s a tme ntensve commodty, an ncrease n the value of tme let Qat consumers feel wealther and more wllng to ncrease the consumpton of a commodty representng a very mportant mean of socal ntegraton. 6. Conclusons The monetary emphass placed on Qat consumpton may be contrasted wth Western atttudes toward drnkng, smokng and other addctve behavours whch focus prmarly on the quantty of cgarettes or drnks a person regularly consumes and secondary on the money cost. The preoccupaton s more wth the consumer s physcal or psychologcal susceptblty to these substances and less on how hs consumpton habts affect or reflect hs economc crcumstances. Wth Qat the emphass s reversed. Ths s 18

19 because Djboutan men know that each person s level of Qat consumpton s related to hs fnancal status more than hs bodly or emotonal needs, and so the more one can consume the more prestge he wll be able to gan at socal level. From the male pont of vew of Qat users, a man who spends modestly on food and a lot on Qat s not necessarly jugglng hs prortes between famly sustenance on the one hand and selfsh pleasure on the other. He may be optng for one prestge system rather than another, possbly because he cannot afford to compete n both choosng ether can be responsble and sensble decson because socal and economc advancement may depend on t (Wer, 1980). The purpose of ths paper was to study how the consumpton of a socally addctve good responds to changes n the value of tme as represented by wages, n a model n whch the Drug experence s produced by usng both market and tme nputs. Estmatng a smultaneous system of equatons wth wage and Qat consumpton as endogenous varables, our fndngs suggest that there s a clear relatonshp between Qat consumpton and the wage level. In partcular, the paper found that the last one postvely and sgnfcantly affect the consumpton of Qat. Gven the realstc assumpton that Qat s a tme ntensve commodty, ths means that ncome effect (generated by a wage ncrease) wll overcome the substtuton effect. 19

20 References Becker, G.(1965), A Theory of the allocaton of Tme, The Economc Journal, 75 (299), Becker, G. and Stgler, G. (1977), De Gustbus Non Est Dsputandum, Amercan Economc Revew, 67, Becker, G. and Murphy, K., (1988), A Theory of Ratonal Addcton, The Journal of Poltcal Economy, 96(4), Becker, G., A Treatse on the Famly, (1981), Harvard Unversty Press. Djbout. Crossroads of the Horn of Afrca. Poverty Assesment, Report No DJI, Document of the World Bank, (1997). Heckman, J. (1979), Sample Selecton Bas as Specfcaton Error, Econometrca, 47(1), Heckman, J. (1974), Shadow Prces, Market Wages and Labor Supply, Econometrca, 42(4), Kaestner, R. (1991), The effect of Illect Drug Use on the wages of Young Adult, Journal of Labor Economcs, 47(3), Kennedy, J. G. (1987), The Flower of Paradse. The Insttutonalzed Use of the Drug Qat n North Yemen, Dordrecht: Redel. Mncer, J. (1974), Schoolng, Experence and Earnngs, New York, NBER. Republque de Djbout: Rapport Natonal sur le Developpement Socal (1995), Mnstère du Plan, de l Amenagement du Terrtore, de l Envronnèment et de la Cooperaton,. Salvn, A. and Zanellato,L. (1998), Pscologa Clnca delle Tosscodpendenze, Lombardo Edtore, Roma, Italy. Salvn, A. Teston, I. and Zampern,A. (2002), Droghe : Tosscofle e Tosscodependenza,UTET Edtore, Torno, Italy. Turch, G.P. (2002), Tosscodpendenza, Domenghn Edtore, Padova, Italy. Wer, S. (1980), Qat n Yemen : Consumpton and Socal Change, Brtsh Museum Publcatons. 20

21 Table1. Mean and std. devaton adult household heads aged Observatons: 1787 Mean Std Dev Household head age Household head schoolng % of alphabetzed % marred % employed %lvng n the captal Djbout Weakly wage (Djboutan currency) Weakly non labor ncome(djboutan currency) Weakly Qat consumpton (grams) % Qateur Source: author elaboraton from EDAM data Table2. Weakly Household Expendture shares Food 0.45 Water 0.03 Educaton 0.02 Rent 0.18 Health 0.02 Clothng 0.03 Transportaton 0.04 Energy 0.07 Qat 0.11 Cgarettes 0.01 Transfers 0.01 Others 0.03 obsrevatons 1787 Source: author elaboraton from EDAM data. 21

22 Table3. Parameter Estmates from 2SLS Qat Consumpton Equaton Varable OLS OLS a Tobt a age (-2.52)* (-1.67)*** (-1.94)** nstructon (-2.16)* (-1.78)*** (-1.99)** alphabetzaton (1.12) (3.24)*** (3.34)* chldren (1.41) (1.89)*** (1.76)*** chldren (1.42) (1.36) (1.21) marred (-0.84)) (0.05) (0.09) Djbout (0.39) (1.20) (1.32) urban (0.80) (1.27) (1.55) ln_wage b (2.20)** (2.65)* (2.62)* non labor ncome (-1.42) (-1.58) (-1.49) F 5.50* 15.04* χ * Note: t- statstcs n parenthess. *sgnfcant at 1%, ** sgnfcant at 5%, *** sgnfcant at 10%. a Selectvty corrected b Margnal effect= and s statcally sgnfcant at 1% 22