Demand for primary schooling in rural Mali : should user fees be increased?

PRHl- ?70-Y PHUN Teclaical Note 87-4 DEMAND FOR PRIMARY SCHOOLING IN RURAL MALI: SHOULD USER FEES BE INCREASED? by Nancy Birdsall Originally issved in Country Policy Department Discussion Paper No. 1983-8, February 1983, "Three Studies on Cost Recovery in Social Sector Projects." Reissued January 1987 Population, Health and Nutrition Department World Bank The World Bank does not accept responsibility for the views expressed.. herein which are those of the author(s) and should not-be attrih'ted to the World Bank or to its affiliated organizations. The.findings, interpretations,. and conclusions are the resalts-,of research supported by the Bank;: they do not necessarily represent official policy'½of t Bank. The designations employed, the present tion of nmaterial4: andaidy maps used in this document are solely for the convenience of the reader and do not imply. the expression of any-.opiniid-whatLoever on tde, part- of the World Bank. cr its affiliates concerning the legal status of any. country, territory, city area, or of its'authorities, or concerningthe delimitations of-its Soundaries, or 'national affiliation. PHN Technical Note 87-4 DEMAND FOR PRIMARY SCHOOLING IN RURAL MALI: SHOULD USER FEES BE INCREASED? ABSTRACT This paper presents estimates of the price elasticity- of demand for primaty schooling using household and school survey data from rural Mali. The elasticity of enrollment with respect to the price, i.e., the local school fee, is compared to the effects on enrollment of distance to the school and various indicators of school quality, including books per classroom and the number of grades offered. This price or school fee elasticity is of interest because a greater degree of cost recovery through fees provides a possible means of reducing the burden on the public budget of expanding and improving the primary 'chool system-admittedly, a second-best solution given that the public subsidy to primary schools is already much lower than that to higher levels of schooling, and the poor are much less likely to benefit fzzm the higher levels. Though fees have a negative effect on enrollment, the results suggest that distance to the school is a much more critical factor. Average village size in the region studied is only about 400, villages are relativt_l dispersed and transportation is poor; thus reducing distance to schools from households would require building and staffing many very small schools and would probably raise unit costs well beyond what modest increases in fees could cover. On the other hand, the results also show that certain improvements in school quality could easily offset in terms of enrollment any negative effect of higher fees to finance such improvements. For example, the number of books in the classroom, an astonishingly low number of two on average, could be doubled for a 10 percent increase in the fee; the point elasticity estimates indicate the net effect on enrollment would be positive. The other indicators of school quality suggest that the central government faces continuing difficulties meeting the recurrent costs of maintaining schools. This, combined with the fact that orimary school fees are assuredly regressive in Mali, means that any case for user fees in rural primary schools must rest on a strong condition: that such fees are controlled at the local level and used for books and other inputs to school quality, rather than chanelled to the center and used Lo red;Ae the government's fiscal burden. **A* *****A Prepared by: Nancy Birdsall This is one in a series of par:s prepared under Research Project 672-72: Demand for and Willingi,iss to Pay for Health and Schooling in Mali. Originally issued in Country Policy Department Discussion Paper No. 1983-8, February 1983, "Three Studies on Cost Recovery in Social Sector Projects." Reissued January 1987 In this paper, the effect of school fees on primary school attendance is measured, using house'old and school survey from a poor rural region of Mali. The elsticity of demand with respect to fees is estimated, and compared to the effects of distance and of quality, including such school- specific quality measures as books per student, the number of grades offered and the recent success rate of students on a standardized test. The question of the effect of school fees on demand for schooling is a critical one. Many developing countries, and particularly those in Africa, are having difficulty financing the expansion of primary schools from wtat are still very low levels of enrollment (in Mali, the primary school enrollment rate is less than 25%). On the one hand, a greater degreQ of cost recovery through direct fees providles a possible means to reduce the burden on the publ:c budget cf such expa.nsion. On the other hand, imposing or increasing fees could reduce demand for schooLing, particularly in rural areas where expansion is most required; in rural areas household incomes tend to be low and in many countries of Africa parents already bear substancial costs of schooling children, at least as a proportion of their income, given the opportunity cost of children's foregone labor both in agriculture and in work at home (e.g. fetching wood z-nd water), and given the costs to parents of transportation and school materials. In this respect, pri=ary school fees are Likely to be regressive, particularly in ali where higher Levels of schooling to which the poor are less likely ever to go, are much more richly subsidized. I/ To permit expansion of a school system, an alternative or companion strategy to ..easing the degree of cost recovery is to reduce unit costs. Though this :!ver is concerned pri=aril! wi:h the first option, I.e. .i:h :he feasibilit7 of increasing user fees, cha e=pirical work also providas so0e indication of the potencial effects on da=a=d of cAr:ain =easures to reduce unic coacs, iz particular of raising che scudent-ceacher racio. 7he paper is organized as follows. In Sec:'on 1, A siple ol of household demand for schooLing is presencad. 7hs model sugaests a fraaework for analy:i:g ..a potential effec:s on demand for school fees. -:phasis is placed on che need to distinguish becween the effec:s on demand oct only of the user fee but of che quali:y of schools and the distance of schools from huusehold. En Section 2 :he survey data and sectizg in rural Mali are described. Informacion from cha household daca on income, and from the school and village data on the cost of scooling to parencs, cogecher provide an indication of the abili:y of households to pay for schooling. 'n Section 3, the household de=and codel is estia:ed and the resuL:int estimates of eLastici:as of demand w7l:h respecc to fees, diszance and various c-mpomencs of quality are discussed. In a concluding section, the implications of the analysis for cost recovery policies in rural Africa are suggested. - 3 - Section 1. School Fees in a Househuld De=and Model Consider a simple model of household demand for schooling. 2/ Parents derive utility from the expected future income of their children, and from the household's consumption of goods in this period and the Leisure time of all household members. Zxpecced income of children is a function of the time they spend in school, the quality of the schools they acteand, the tie they spend in ocher learning oe skill formation activities outside of school and a see of child-specific characteriztics which affect the child's recurn to schooling and to other skill formation activitias. Thus schooling is viewed as an invest:ent good. Parents face a tradeoff between household consumption now and children's expected future income. T-e cradeoff occurs because schooling is not free to the household. The= are direct costs (fees, transportation, books); and there is the indirect cost of children's foregone Labor in paid work or at home (where they may substitute for parents) and their foregone leisure. !a addition, there is a tradeoff between time children spend in school and time in alternative learning activities; if the latter can be combined with work in the home, ii a family business or on the.family farm, time spent in school will be more coscly even given &a equivalent return in future expected income, than ti=e spent learning how to farm or cook. The household, in other words, faces a budget constraint; total spending of =oney and time cannot exceed the sun of earned and unearned income. Schooling of children has both money and tie costs. Viewed in this way, demand for schooling of children is analagous :a demand for any good and can be created in a conventional consumer demand framework (eyen chough schooling is for the household ac:ually an ivesment). \he quanc±:y demanded of schooling will be a function ot 027 of household incoe (income affacting spending on this investmenc (assumi:g capital markacs are imperfect and parents cannot borrow against their children's future Income), the prices of all nonschool goods, and the price of schooling in ta=s of user fees, but also of ocher factors that affect the costs and return to schooling from the point of view of the household. Zz addition to direct user fees, these include on the cost side: - costs associaced wi-h the distance of the household from the school, inclu&ilag the opportunity cost of travel ti and direc: costs of any cr::=port; - the opportunity costs of the child's time foregone from vork, leisure and other learning ac:ivities; and on the returns side: - the quali:7 of the schootin3 available (for a given amount of ti= in school, lower quality yields a 1o%er expec:d Income recurn); - the characteristics of the child (ability, sex) which affect expecced returns. The idea of the =odel and Its applicacioa t :he user fee issue Is illustrated In %aures 1 and 2. In Fiure 1, 1.' the horizonalaxis, Q, represents the quancicy demanded of schooling, and the vertical axis, ?, the price (or fee charged). Ds Is the social de:and for schooting, D the pri:aca demand (wi:h the difference due :a positive ex:arnali-es and,'or -:perfec: narkat zonditions I), and Pc the marginal cost of provii- schooling. SS Is a locus of points showing the quantity of the service the governme: can prc-Lde at different levels of user fees. (The tzportant assumption is tha: the &overnment budge: for schooling Is fixed and that fee Increases ara used 斗r&！他 一一一，〞－-~-－一－j•－-.－網•自網•闕－向。．■．,.-－網調•••．．••．••■‘．•• 嚇‘-．守、合寺州· 卜一一付一一一V一一一一弋一仕州。。 筆頸’& .,O.4»vncv lrýz c: .29$0-,D pp?,^cze R:lLm 9:,J.ZIM ring IGDJfavs Imlef P---- zc; flojprno-ý =tt wý2 ;c, ML_:r.z Pule--ap N: zc IUMT.:r 67 &Ljnz IPDO.395 t IBZTJ- 1:, 1: t nb S: I Aq pur I wv ;kj et T 2=;z7T;zn KZ;r. Panr;Dover £2rbz uCT3 pur 1a4=2 ^2 Aq znq IP92,znz &va; iq £:=D not PPV7TZ:vZOD el purmap'ozlrATzd 3n2 Top= er:;: låel; jlvzri .933£29% :o a--nzvc Irg: rl ssulmp T P9277.>OSST 6.; :,, ;-, - ---, ovv32213, ljum p--? - purzwp not pavv vz-.4:ws r ao; uz;ad r-, wsceazr3 Irf - azvdc 20 =,. juko Spoos onovmoscroN za^ut. IT-Lenza.JA alt sr-,Tool;zf pur lz.ýllvwq VY7-1 sý-0-.zeg U:7 -P-~ P LTIMIon TT7^ ¥Z;Ljgs 20 poos r*.4;f T zo; DzTje t7 usegazzi av 2v1;2 ur_,z 67 21 i4:tjenbwc, 9nPclza-jyxa pTnor. *a; r-, arvaazrj Aur zvý2 os llood v:;: ä~ Lla%7zaid zsvvraz£ qq c2 ÅTn71 -.%ro-- nor no-pd~co Tmpdoqne s7%;. s-, lao^oeacrg joo; aucn r jo; T-,r zT owTz om eanz uwgz sj Gap; Jorn av; DWVD uu Ab o: lac-:jn; TT7ts no-gd=wuoz emrnpaj lv=rs V-P SeCTq2 Jor30 TTr IPTnoän Ii 02 Arc IP&22r4D ge; n2 VI ellvic=I Årv :r42 219TZ 67 2T Ipziir--ep 66*2x9 zuaavddv cm lamemop -Bzrm*Tzvz ;c- ocabap n: ovnpej pur uvT&im unz p=vdxa on z-ao=aAoS zT=od pInoz p92avip ozTad ir42 =1 «wazzul ur 192=vqj ot zir uzV_zwc n2 ;o qgnon* vpiacad on ;zo;;? 2= PIncD PtTOWP gsvm%w alen 92DRZ ;j -drz-n =;jq7-,Tnbw -Iab;bj-.nt>l jo pr-ý t cl UoTzvr.n,.r> !PUTM9P gr.93x9 .10D Alddns 9692x9 .2n,,;vv 11DAIL-IC4 16, cicu *~2n0 ut pur =c,2d=CZCDJWP= rnr. vi funi logarn) ou %p". ubv: * Ob r-, rc-,zd=wuoz ;o ::=no= T=_,.:do 9-,2 limuncn -pcw-wp gr-poom nernr 'a2.2vqz cz :v up;Acäd rvz 2-jo- o.4o2 irý2 Imnwp vy -wezsXs 9-ý,å puedx* 03 - 9 - -7- Nore that, as drawn here, the quaucity demanded of the closer service, Q2, is greater, even with the fee, than the quantity demanded of the free, but more distant service (Q-,). With higher quality or lower average distance, there may of co-.rse be an increase In mar,;inal costs (unless these can be obtained through more efficient allocation of resources), shown as the new marginal cost Line, c, and resulting Ic a shift upvard of the SS curve. Obviously, the less marginal costs increase, the greater the net increased social benafit. 5/ The incorporation of distance and quality as factors along wi:h fees in govreraing demand for servicns is especially important in analyzing demand in poor rural areas. The region of Mali considered below is only one example of a simastion where there is no apparent excess demand - for health, water, schooling, agricultural. extension, even for roads - and where a kind of low- leel equilibrium trap e.'scs 6/ such that expansion of a service seems unjuscified. Iz these si:uacions, the question of the extent to which distance and quality are factors .1miting demand obviousl7 must be asked. Suppose, houever, chat the governmenc cannot afford to reduce distance and improve juality, under current and expected future budgetar7 constraints. Then the question arises whether user fees can be '.posed to help pay for expansion and Lmprovemenc, and what effect such fees will have on demand. Referring to Figure 2, it is necessary to know: - the amount by ahich the demand curve shifts outward with an imp,cvement in quality or a reduction in distance, i.e. che elastici:y of demand with respect to these factors, - the slope of the demand cur-ve, i.e. the elasticLty of demand with respect to a fee; and -8- - the relation beceaen fees and the increment in the nargi=al costs of an exanded and improved system, i.e. the extent to which an increase iv !ass could generate revenue sufficlenc to pay for services that are closer to households and/or of better quality. :/ Obviously If distance and quality =ac:sr little or not at all, the shift in the demand curve shown in Figure 2 will be minimal, and a fee will reduce demand below i:s level. without anj fee. Similarly if the slope of t!s dama. curve is not at all steep, fee increases will have a relatively strong agaci,e effect on demand. 'Jill fees exclude the ooor? To the extent that the poor already have access to services of reasonable quality, an increase in fee may well reduce chair demand. However in the region of Mall discussed here, and ch.:oughout =ich of rural Africa, the utilization of available schooling, as shall be demonstra:ed below, is clearly limited by discatce, as %mll zs by the inabili:y of schools to provide a reasonable eavtroacena for laarning because of 'Ack of ceaching =acerta:s. As noted above, :bese factors Imply private costs, direct and indirect, which may exceed in mgitude the costs due to actual fees. To the extent that residencs of such Liral areas are all relatively poor, at least i= the sense of being currencl7 deprived of nearby services of reasonably good qualL:y, :he same three pieces of information noted above (elasticities of deand with respect to distance and quality; with respect to fee; and relation betveen fees and costs) are all cha: is relevant. 8, 1: is even possible :ha: :he *3verall cost of utillza._on could fall vich a !.e if the service vers prov!.ded closer. The issue of course t1 more complicated in urban areas, where be::er -9- services are available to the urban poor. There it is likely that though te urban rich benefit more than the poor when no fees are charged, since they use the free services more; stil.L the urhan poor do at least have access, and imposition of fees would for them clearly raise their overall costs of utilization. But would affordable f:-e- cover ccsts? Regarding costs, the model in principle requires that -a ask whether user fees could cover the iccrease in long-run marginal costs associa..d with higher quality or a more widespread system (though not, it should be pointad out, whether fees could cover total costs including capital inveitment costs.) In fact, under many circustances, including the situation in Mali discussed below, the critical question is whether fees could cover an amount somewhat lower than long-run marginal costs, i.e. the marginal costs of inputs such as blackboards, books aa& various teaching materials that are complementary to but do not .nclude the salaries of tiachers. The difference is important; in Mali !or example, the overall budget for a typical province allocates 93Z to personnel costs and 7Z to complementary inputs such as supplies. 9/ As a practical matter, it is often these complementary inputs :hat are shortchanged in the face of a central governmer: spending constraint; it is usually very difficult for government to reduce the worklorce in zhe short run, and even in the long run given political const.derations. The result can be a dramatic decline in the quality of the service. In the case of the survey region of Mali, for exampLe,' the survey data indicate at least tvo ways in which the goverment budget problem affects the rural school system: firse, teachers are regularly paid as much as three or fcur months laca; and second, :he araIab.ity of tex:bocks is shockiagl7 low: 6 e :a00 students, or only 2.5 backs per :-Lass. :- fact seven of the 20 schools surveyed had no books ac all. 1O/ User fees at the low levels affordable in rural ?ali could not resolve the problem of late salary pay*nts; but a useful question is whether thAy could te used to assure that books are available in ever7 classrooc. - 11 - Section 2. Survey Data, Setti'g, Private Costs of School The data analyzed below were collected in tvo related surveys conducted in the Kayes region of southwe3tcrn Mali: a survey of 186 households Ia 38 villages conducted in April 1981; and a survey of small groups of men and woman in the 30 villages and of employees at schools and health poscs serving those villages, conducted in Narch of 1982. The villages included in the sample.were chcsen partly on the basis of water scarcity; households wiChin villages were chosen at random. 11/ The household data include staudard socioeconomic and demographic information (household income and sources of income, farm productior data, age, sex, occupation of household members and information on the number of persons enrolled in school and the number of adulto who a.e Literate.) The principal purpose of the village and school survey was to obtain information on fees, school quality, and on the distance from survey villages to the closest schools. These data were matched to the household data, so that the parameters of the demand model outlined in Section 1 could be estimatee. The 38 villages where the survey took place are located ia three contiguous administrative districts, or "cercles", of the Kayes region: Bafoulabe, *Kanieba and ita. 'he three cercles, and especially Knieba, are in a rugged and remote region in Mali. Rainfall is confined to a four-=onth rainy season; transport is extremely difficult at all times and for some villages impossible during the rainy season. The villages themselves are small, with about 800 persons, and are on average 50 km from the nearest market town. ?er capita income is less than S200, 12/ making it one of the - 12 - poorest areas ta :he world. Life expectancy ac birth is onl7 32 years. 13/ The majority of the population is engaged in subsistence agrtcul:ure. Educational status is also extremely low. Only 8- of the adults in the surveyed households reportad being literate (the adult literacy race ia Mall is less than 10%). 3ased on the number of school enrollees in the survev villages and estiaces of the populacion.in those villages aged 6 to 14, :he enrollment rate is probably below 20w, somewhat lower than the ra:a of 21: for Xali as a whole. (The gross enroll:ant ratio, which includes underage and overage students in the numeracor is 30% for Mali; fcr households in our sample, it is even higher,44Z, probably because i: includes children living in ocher villages in order co attend school and not included ii the denominacor, and because some respondents may have reported persons who were enrolled in the past to be currently enrolled.) Ability to ?ay schooling costs Given the low level of izcome ia the survey rsglon, a firsc Issue ia the amount of costs pareas currently incur when they send children to school; wiat proportion of annual household income is required to send ore child -or ,,e year? L. Direct costs 14/ All prizary schools ia the sur-vey region are public schools, which nave no official fees. 7ovever, the school survey indIcated :vo :yes of imformal fee- paid by parents. 7ees :o parent associations -wera cn average 708 MF per year (about S7.5). First-cime enrollment p:-s concinuing karolLnent fees ranged from about 60 MF :o 250 %T per year (depending on :he arumpcion made regardi=g whether reported fees were ore-:za-only or annual). Cur survey resul:.% ngarding parent associa:ion fees indiza:a - 13 - amounts collected in this region are far below the amount the governrent assues is being contributec . A review of education in hali done by the government wl.th UNESCO suggests that contributions per student to parent associations should be 3500 F, compared to the average of 708 MF in cur survey. 15/ ased on infor.ation provided by school headmasters and 4nterviewers about the quantity of notebooks, pens, and slates studeuts are required to have for use at school, and the prices of these goods, we estimate total yearly expenditures on nupplies per child per year of primary school to be about 2,900 MF (about $). 'We have no measure of direct transportation costs. Virtually all children who attend walk to school in their own or a nearby village, or live with other families in -rillages that have schools. There is do public transportation system - indeed La some parts of the year roads are impossible. (?uel, mechanical skills, and vehicle parts are scarce in the region, due to nonexistent di. .bution systems, so that the idea of a school transportation system to reduce the effective distance to schools is probably not practicable.) Thus for parents, the major component of transportation costs is the cost of the children's time going to and from school. Distance is a partial measure of this cost not only because of travel tize but because beyond 10 kilometers it is impossible for children to valk, so they must be sent to live in other villages to attend school, and cannot work at home at all. In the empirical analysis below, the two measures of cost used are distance and fees. The estimates of required spending on school supplies do not vary across households as do distance and fees, so they aaanoc be - 11. - expLici:ly included in the de=and ea.ua:ion. it is important to coca, ho.aver, that che the fee represents a sall component of the tocal costs of schooling to paren:s. Of direct onatar7 costs it is only 25Z wi:h school supplies representing the ocher 75%. And the opportunity cost of foregone labor is probably more important than are the monecary costs. Total mnetary costs, including fees and supplies, &e estimate to be about 3850 *F (7) annually. This turns out to be a very small percancage of average household income-less than 0.%. Given that most households are likely to have several children of school age in any one year, primary schooling could absorb 1.0: of household income annually. A useful .comparison is that households in the sanple spend on average almost 5% o! es:imaced income annually on coder- and tradi:ional drugs and about 50: on food. L6/ 2. Opportunit7 costs Our data provide no basis for estimani a shadow prie of child labor. Howveer, information collectad on :he activities of school children after school suggest the shadow price is relatively low during =ost of the school year. The peak periods of de:and for agricultural labor are ia Apr!.- May (land preparatioa) and September-October (harvest). During the rest of the school year, betveen October and April, labor da=nd is low. In 22 (cf Z3 villages reporting), villager respondent groups reported that boys spent ions time tn recreacion after school. For girls, this *das the case tn 13 (ouc of 23 villages reporting). :n no villages wre girls reported co wark in the fields; in 2- of 23 villages they did housework after school. The implica:icn is that agricultural work is not a najor activity of children and chus =o: a major cause of nonenrollment-though help at hom& :ay reduce enrollent cf gIrls. (The responses of villagsrs on a::-.de ques:ions provide no beCtr - 15 - indication of the importance of opportunity costs: the need for children to work at home was cited as the mos: important reason for nonenroll=ent in only 20% of villages, but as one reason among others in the majority, 65%, of villages.) Because we have no simple proxy for the opportunity cost of child labor, we cannot quantify its importance relative to other factors. 'o the extent it does matter, i: is Likely that adjustmenca in the school schedule to free children during periods of peak demand agricultural Labor would largely eliminate opportunity cost as a barrier. Unfortunately, as shall be seen below, other barriers, particularly distance, are more difficult and costly to eliminate. Is there excess demand? Finally it is worth clarifying that excess demand and possible resultant rationing are not an issue in this region, so that the model outlined in Section 1 is particularly appropriate. There are two useful pieces of evidence. First, class size is relatively low - 35 studeats per class compared to an average fo- the country as a whole of 46. Second, in not one of the 38 villages did respondents indicate that lack of space was a reason for children not enrolling, even when directly asked. 17/. - 16 - Seccion 3. Z=irical lesul:s Table I shows the results of various regressions in which a dependent variable representing a household's demand for formal schooling Is estimated as a :unc:ion of household income, and the "price" variables of the model outlined in Sec:ion 1: distance to the school, school quali:y neasures and the amount of the parent association contribution parents -ra expected to pay. The definition of variables and their means and standard deviations are provided in Appendi. A. The dependent variable is the difference betreen (columns 1, Z, 3) or the ratio of (column 4) the number of persons in the hnusehold currenT:Z7 enrolled in school and the number of persons in the household bet-jeen the ages of 6 and 14. 7obit rather than ord!nary least squares is used in column 4 where the dependent variables cannoc fall below zero. The school quali:7, fee and distance variables are the same for households wichir: vilIages; they refar to the pr!.ar7 school which is closesc to the villagQ uhAre a household Is located. 7rfor:una:aly the househ:1d da" do not show which household members were enrolled but cnly the torAl enrolled; thus *e cannot be sure the numerator does act incltude some persons under age 6 or over age 14, and te cannot control !or age and sex. (Ze do cnov from :'-e schoo'. surey t:: atil? 24: of the s:udent bod7 for first and second cycles ara girs, on.y !. of teachers are female.) Thus -:hA dependent variables are relati-i.ly crude aeasures of householm schooling demand. - 17 - The effects of variables as seen in the regressions are brief4y summarized below one at a time; then their effects relacive to-each other are considered, using elasticities derived from the regression results. The efeect of fees Fees (i.e. the size of the expected contribution to the parent association) have a negative effect on attendance, as expected, but the coefficients are aot statist!.ally significant except in column 3, where a dumy variable indicating whether any fee at all is charged is also included. The dum..y variable was included because interviewers reported that parent "contributions" were rather arbitrarily set by school directors, and that in villages where demand for schooling was very low, directors set no fee at all. The use of this variable does show that the existence of a fee is posit'vely associated with enrollment (causality cbviously not being from the existence of a fee to enrollent); but the results also show that the size of the fee has a negative effect, once its existence is taken into account. 18/ As noted above, fees in this region represent oml7 a small portion of the tocal =onetary costs of school, *o it is not surprising that their impact on attendance, though negative, is not terribly robust from the point of view of statistical significance. 'When villager respondent groups wers Asked whether a reason for -onecrolIent was that schools are too costly, they answered an In every village. At only 2 of 21 schools did school headmasters believe that costs were a factor helping to explain low enrollment. The effect of distance Distance to school has the expected negative effect on demand; the rigns on the variables are 'tatisticall7 significant in all 4 col%=ns. Oumy variables are wotd in colu=n 3 to capture whether distance effacts are - 13 - nonlinear. Te dusy vartables suggesc cha: in fac: the negacive effUc: of discance Lncreases at a decreasing race; this ls probably because the cuscj= exists of sending children C3 Live outside cheir own village if no school is avallable tn the vtilage. .us bayond a certaln distanca, or kllo=e:crs natrar relartvely -ikle. A.s wih the fe results, the epirIcal results on di4stance are consisan -ith the resulzs from the village at:c:de survey; Li abouc ona-half the villages, the fact that the school was far away was reported to be a cause of =onenroll.ment. Notc that with :he use of the du=y variables, the absoluta agui:udå of the coefficlent on :he fee drops (compare colun 2 ro colum-s I and 3.) Effects of school cualit7 There is a Larg Lterature on the effects of vartous measures of school qual1:y on student ach-leve=enc, e.g. tesc scores. 19/ As suggested tn the Section 1 nodel, berter school quality is lkely to increase enrolinenc as -all as achilavement of those enrolled becausa it increases the recc-rs to schooling for a given a=ount of :=e and direc: costs. in the reg±in of Ma.± under study, t: is possibla thac better school qual1:7 tncreases recur~s and ra'ses enrolltenr si±ply by nceasi=g tha lkllihood that chtlåren, once enrolled, will sray enrolled, rather than drop ouc because they are not LcartUng anything. = ch* village sur-my the prt=ary reason sug3estad for norenroli.ent was :-a: chldren "dislie school" (ci:ad =ora of!ar. tan :-a ned for ch.ildren ro work, or tha: the school was too dis:anc. :n the sur:ey of school personnel, the "difficulcy" of school vork for children was clzad by 18 of 21 respondents as a reason for children diopping ou= (cc=pared :o 7 vno cited the necd for children :o work at home) - 19 - The table in Appeudix A includes the mean -ralues of our school quality variables. Though teacher education is not high, at an average of 11 years it would be adequate if teachers were provided effective support in the form of textbooks and teaching aids. However, the table suggests such support is very Limited: seven of 21 schools surveyed had no books at all; (not shown is that four did not even have chalk). It is not surprising that children have difficulty when the lack of books and other teaching material, (as well as convencional teacher training methods), means that teachers conduc: classes in which the main activities are copying into workbooks or onto slates what thie teacher puts on the blackboard, and recitation by rotc. The quality indicators show that only 39. of students in this sample who took it in 1981 passed thc primary school leaving exam which qualifies them for secondary school (60% in the country as a whole). Teachers are on average of 10 weeks behind in receiving salary payments. Most schools do not have 6 grades; a large number accept new students only every tvo years, so in any given year offer only grades 1, 3, 5 or: 2, 4, 6. The student-teacher ratio of about 35 is noc :oo high; but thii is more likely due to the dispersed population than to any purposeful effort to keep the ratio low. Not shown is that 37% of students in the schools surveyed were repeaters (27% in Mali as a whole). 7able I indicates that on the whole the varicui =easures of school quality do affect attendance, in the sense of having a statistically significant effect. Teacher education, the variance of which is -try low, has no effect (it is close to significant but negative in column 3). The student- teacher ratio does =a::er; this is unfortunate from a cost point of view since it suggests that were other s..nditions to increase enrollment raas in existing schools, econcoies of scala could not be realized and new classroo.S and teachers would have to be added. *'hether a school has 5 compleca 3=2des .oes not affect enrollzenc, nor does the pass rate on the school-leaving e3az. Roever, the num!er of books per classrooM and the lace pay=ent of teachers (perhaps reflec:tng the general level of admimistractve sun-ort teachers can expect) both have the expected effec:s. Given the smal size of the sample and the Lack of variancs a=ong chase schools in quali:y, the results are quice robust in the sense of indicating the effect school quali:y can have. At the same ci=e given asu.r:ujlLnearity among :he masures, not too much should be =ade of the effect of any one measure as opposed to another. Non-price effects Interestingly the effect of household incmes appears to be cagat4ve when the ratio of adults to children is included (columns 1 and 3). It is possiblA this reflec:s the contribution of children moc in school to inco=e; because zeither our work nor enroll:ent data are s;ecific to lndividuals we cannot tell. Ethnic and religious (the exten: of Islaz) differencas across households had no effect on demand (not sh-w) once ocher fac:ors were taken into account. There are loramic schools in che region; however thair hours are usust.y short and La the evening and fieldworitars reported :he7 are =no ca any may viewed as an al:ornative to formal s.:hoolhg. 'ndeed some households had persons enrolled in both forma and Koranic schools. There were also no significant differences across :!e three carcles. 3ecause the pr±zar7 econozic activit in a !aba is god- in a1 which child7en can .:e %lpful all 7ear rtiid, we ancl:;a:ad a lz:er decand - 21 - there, cater.s paribus. The lack of any difference underlines the possibili-7 that opportunit7 cost is not the major factor limiting demand.- Elasticities of demand Table 2 shows point elasciiies of demand vith respect to fees, distance and selected :easures of quality, calculated on the basis of the regression results in 7able 1. The columns in Table 2 correspond to those in Table 1. The elasticities allow rough comparison of the relative importance of the different effects. The ns notation indicates that the elasticity was calculated on the basis of a coefficient whose sign was not statistically significant at the 10% level at least. Table 2 indicate3 that the importance of fees relative to distance and the quality easures is hi-hly sensitive to the particular specification. where it is scatistically signi!icant (cols 1 and 3), the negative effect of the fee is greatr in absolute terms than that of distance. If fees were raised at the village level and there war no change in the distance to or quality of schools (i.e., the funds were used to reduct the current overall fiscal burden of the government), the low enroll=eat rates would fall further. Given the reLaYCvely low fee level, moreover, it is unlikely that fees could be usec& to finance the creation of more schools in order to reduce the average distance to school. Though tocal enroll:ent would increase if schools were available in =ore villages, i': is not clear that enroll=anr increases would be great enough to keep uni: costs fro= rising, particularly given that villages in the regica are sall-average population size is only about 400. indeed it is possible that enrollenc in some existing schools would fall, as children who now attend from outside :he village shifted to new schools ne.arer to their homes. Siilarly faes could moc fiance the e:ra costs :ha: would te associated with reducin3 :he studenc-teacher ratio. From this point of view alone, the situation is rather grim. : costs are bound to be high where the population is highly-dispersed; the situation is more difficul: still where demand for schooling is so senitive to the distanca factor and where transportation i-s poor. Rowever, comparison of the relative effects of the books per class variable and the fee indicates that there is so=e scope for increasing enrollment without adding to the fiscal burden of the cenctral gover::an:. Though the point elasticit4 of demand with respect to books is lower in absolute term than that of fees, a small increase in fees could probably provide for a dramatic increase in the number of books per classrooo. At an average of slighcly cver 2 books per classrcom, it is obvious that for soma subjects teachers curratly have no back at all even to guide them 'n planning instruction. Iadeed the lack of books and other teaching macarials =&y explain the negative effect of the student-ceacher rztio, which is czherwise surprising given the relatively scallI class size. .eachers =a* he aa effec:ivA-though very cost7--subsci:uce f!or :acerials. L'i:h 34 chi:dren in a classroom, an increase of 10: in the average fee (from abou: 700 SF to 770 S) would produce 2380 MF (about S4.50). If books coa: 1000 MF the average num:e: of books in a classzoo could be increased A0C for a 10Z tzcrease La fee. (This of course assumes that an effectlve sys:am for producing and Aistribucing bookj exists, and that any increment in current fos is used f!r purchase of books for local use, not channelled to :he center to reduce education budge: defIcirs. These are strong caveats,) - 23 - On the face of it, te negative effect of :ne late pay=enc of teachers could be rectified even more easily at virtually ao cost at all. In fact, it apparently reflects not only a gen6ral budget problem of the central government, but administrative and logistical problems which are greater in some local areas than others. 4e do not know what these might be. Hovever it is possible :hat !ees collected and, more importantly, ontrolled at the local level, might help provide for teachers a les- uncertain working environ=ect and thereby for children a better learning environment. Conclusi.: Cosc ?,ecover' - Rura. ria: Prizarr Scoos I. is al.ays dargerous to gaeeralice !-oz a szall sarpla .: -.a: : y be an unrepraseciative area. 'dowaver, i: is prebable that c . a,ve resul:s do have implicaclsas fozr prary school financing i -. ck-, rural rgiSns of Africa, wtrv ch ppuli.rloa is '44ly-disparsed, icoied aza low, and priary The resul:s L-:4ccacs chat iLposig ar I=cre.si=g !ees for schoolig would, all oter thi=gs cha same, raduc. demaad. This is :'.e case evea chough tha fUe per sa r*preseans a :.=all propor:ion of tha privaca cos:s of schooling-in this region at "st 25 perceac of direct. monacar7 cascs, and an tven smaller percenaga of :ha cocal csts, -;ich iclude foragore child labor. Zncoms is very lcw in tha region rd despite de jure cc=pulsor! school attendance, tha ear31ent rac.4 in pri=ry school is less han 30 perean-. '".a qestion cauld justifi.ably be asked whechar chera shouid be any fai at a--- (actua'ly the "fte" zka2s Che for Of an expec:ed cOncribu:0n to the paren: asåocat:lon.,) Ca re cae hand, our asti4=aca of :ha annual dirac: cosrs of school!ng cce child amoun-s :o a small proportia of averag•: hot se'old i:co=e, Oua-t%af cf aa percenc. O the other 'aud, many households wii!ave :-.o or evaa =ore childran of school age, and even one percenrt can be a sustanrtial portion Of Cotal incom given Chat C::a propor4o1 .penc Ga f!od =ay excaed 75 Waeenc'.n soa ycea:s. 2 z/ ro an egt.ity paL of vaw, cha =se cr :ees zs also -jeak: s,all s%4f:s in canc-.a goaerear speading o educa:ion, a;ay :: stipenda for untversi:y scudents -wh£zh are extraord±iarily Ze=rous bu: benefit very fav--nd coward prisar7 schools, could easily =aka eliminacisn f pri=:ar7 school faes possbl,e. (Cniers:ty sci;ends in 1978 :ock up aou: 1W ;erceac of the cierral galernnt eduzaton budger, copared ro 22 .arze=: !:- - 25 - prizary educalion. Unics costs -.ere $55 per primary student and j1,258 pe: university student. 21/) In this context, is there any case for raising user fees? The empirical work. above suggests there may be, at least in a second-best world. First, note that fees are rot a major cause of nonearollment or dropping o. ,. hA emtirical results indicate th.t distance to school is a much more critical factor; the quality of zh. school wlso matters. These quantlcative results are backed by the responses to attitudinal questions in the villages--cost was not felt to be a factor causing noranrnllment. Second, note that poor school quality does reduce enrollment; the quantitacive evidence for this is complemenced by actitudinal daca indicating there is a "diszouragemant" factor--many children find the work too di.ficult and drop out. Sime may rAver be enrolled because parents see little evidence that children can complete school. A case for user fees rests on the possibility that such fees could be used to increase local control over quality, for exampL by enabling communities to directly purchase books. Tn a situation where the central government faces continuing difficulties meecing its obligations (such that teachers ia our sample we-e receiving salary payments on average 10 ueeks late), inputs such as books have been shortchanged. Though such inputs would represent in general only a s=all fracticn of total recurrent costs compared to salaries, it is usually very difticuL. for government to reduce the workforce in the short run. Thus these compleentary inputs are the fi:st to be cut, with resulting problems of poor quality at the classroom level. Even local financing ca-not itsel! resolvc th problem of productica a.ad distribution of schoolbooks. Nor is it clear how such local funds would - 26 - be adminiscared within comunities. 22/ Eo-ever, fees at least provide joze degree of local leverage, and a point of departure for 3rea:er local interest in the educational system. 23/ tes should not be mandated from the center. At the sa-e time local communities should not be discouraged from initiating fees for their own use. Indeed school directors and teachers could be given clear guidelines on the range of amounts of acceptable fees, and craining in how to work with village leaders in their administraclon and use. In this region of all, fees can provide only a partial solution ta th.Q quality problem, and even than jn!y given an adequate distribution system for books. They cannot solve the problem of the high unit costs the: would be entailed to provide schools closer to children's homes, given the degree of population dispersion and the Lack of roads and transport. Th, s me is likely to be trze elsewhere in rural Africa, particularly in the Sahel. The distance problem is not easily solved; but fes provide an entry point for actacking a severe quali.y problca In rural primary schools. - 27 - Table 1: REVUMOW. DV.AnD 701 SCL00".13C UCR-75SI0NS (t-stacstcr in pareucneges) epenent variable: epenunc varable: ~b~r of persons enrollöd minus ber of persons turolled umb,er of persons aged 6-14 (OU.) divided by persons 6-14 (Tob-:) (1) (2) (3)() Rousehold variables: .aUural logar:n -.378 -.126 -.459 .0278 of houaehold income (2.27) (.62) (-2.59) (0.345) Sber of persons aged .622 - .651 13 to 50 divided (6.52) (6.63) by persons aged 6 to 14 School cual' variables: Aversg. educailon -2.2) .622 -4.07 .170 of teachers (-1.JO) (.32) (1.94) (0.216) Srndsnc-teacher -.036 -.0516 -.0804 -.0122 rara (-1.44) (-1.72) (2.10) (-1.37) D y: school offers .319 .0521 -.0227 .1W 6 complet gradez (1.73) (.247) (.089) (1.55) ~uber of books per .108 .0924 .118 .0344 cssroom (1.99) (1.35) (2.17) (1.2) ber of veeks since -.0871 -.0885 -.138 -.0129 teachers paid (-1.84) (-1.57) (2.02) (-.549) Proportion of sudmnts - - -.0477- passing scan"ard test (.055) School Uc. variables: Natural Logarichm of -1.16 -.343 -1.95 -.0295 feet ae parent (-1.75) (-.405) (-2.42) (-.106) association al D y fe. is Zharged - - .984 at clOsest school (1.77) Distance to school variables: xilome=ars from village -.0766 - -.0558 -.04.2 center to 2a.rest (-3.22) (2.16) (3.4) priary school a/ Duy: school locaced - -1.26 aueside village but wichi (-.01) 5 kilometers D~y: school 6-10 --.369 kilometers away (-.579) Dy: school 11-15 - -1.26 kiloeters avay (-1.93) Du~: school 16-20 - -2.39 kiloeters away (-2.52) Dumy: school mor -2.81 than 20 kilometers away - (-3.49) - Constatc 28.4 . -4.26 53.8 -1.51 12 .35 .12 .37 sigma- .573 (7.0:) 3 123 123 123 123 al Maasured at village level based o closesc school. - 23 - 'able 2: MASTIC'MES OF DAM4D FOR SC2COLNG (1) (2) (3) (4) With respect to: iees a/ -.58 -.17 ' -.98 -.016as distance -.34 - -.25 -.18 quality: books per class .12 .11s .14 .040s student-caacher ratio -.65 -.47" -.73 -.068s anuber of weeks since teachers paid -&46 -.47 as -.73 -.068as a/ This is the elasticity of the fee itself (not of the natural Logari:hm of the fee), i.e. the coefficient multiplied by the inverse of the mean of the indepedent variable. as indicates coefficiant in 7able 1 .as cet significant at 10: Lavel at least. - 29 - FOOTNOTES 1/ As noted in the text below (pp. 24-25) small shifts in central government expenditures from university to primary schools could "purchase" more rapid expansion of the latter than relatively low fees which could be burdensome to poor households. Ax the least this issue should be looked at simul:aneously, as part of an overaIl itrategy ot a more rational pricing/subsidy approach in education. For a relevant discussion of the %elfare implications of the allocation of a fixed government budget levels of schooling, see Psacharopoulos, 1977. 2/ The model is based on an approach suggested by Becker, 1963. 3/ Figure is due to Thobani (1982), though modified to show no excess demand. He is concerned with tha situation where excess demand is observed. In the rural Mali-case, there is no observed excess demand for.schooling. For a fuller discussicn of the issue when excess demand is co: observed, see Birdsall (1982). 4/ Without such a difference between private and social demand, the Lost efficient price to charge would be the marginal cost. In the case of education, imperfect capital markets provide the most obvious argument for such a divergence. Poor families are unlikely to be able to borrow to finance children's education. 5/ Note that if the social demand curve also shifts ouward (e.g. because of higher quality schools), the net increased social benefic is also greater. 6/ Clearly the govern=ent could increaze demand be reducing price still further to a negative amount, i.e. a subsidy. The question is whether goverrment could afford it. Subsidies have been provided for some services, most obviously education, particularly higher education, and family planning services (e.g. persons who adergo sterilization are given cash or other goods). 7/ As drawn, the assumption is tc marginal costs are neither rising nor falling. As services like uacion are extended to more and more remote areas, marginal cot could well rise; we know surprisingly little aboun tL .&:h of costs tinder different circumstances. However, ac Icast in this region of Mali, substantial improvements in qu-.iy are probably possible without any increase in marginal cost, as iiscussed more in the text below. 8/ For a more detailed discussion, see 3irdsall (1982). 9/ See Gray and .artans (1982). -30- 10/ For a fuller discussicn, see projec: paper prepared by Martha Ainsworth, `The Demar for Heal:h and Schooling in Y.al: Rasul:s of the Community and Service Provider Survey". 11/ For full documentacion if the surveys, see "The Y.ali Household, Coamunity ;,nd Service Provider Survey: Docu=antaton" World 3ank, Country ?olicy Department, April 1983. 12/ See Ainsworth, Orivel and Chuhan, 1983, an accompanying paper in chis series, for more detail on household Income and expenditures. 13/ Ecols Nationale de M&dicine ..., 1982. 14/ The sec:ions on costs rely heavily on Ainsvorth (pp. 21-28, p. "A and Appendix .1), where a detailed basis for the estimates is provided. 15/ See Rapublique du eali/UNESCO, p. 26. It is not clear vhecher this amount of 3500 IF is an actual average or an objective from the :ex: of the document; thec it is an objective is more likely, given our findings. LL every stu4enc paid 3500 MF, the total private contribution would amount to 20% of the public budget for prizary education. In fact, even the 708 MF (reported by headasters) =ay eaggerace actual paymeacs; a comparison of responses from our village survey and our school survey indicates that underpaymen: is widespread; villagers raported a mean annual fee of 270 MF,- less than half the 708 reported by headmasters. And osc c the villager respondent groups cice lack of funds as a :ajor problem of :he pare-t associations. 16/ irdsall and Chuhan for drug expenditures, Alasuorth ec. a!. for food. 17/ See Ainsworth, pp. 44 to 46 for docuentacion of villiagers' responses to this and oc'rer at:itue questions referred to below regarding reasons for not enrotL.ig and dropping out. 18/ The possibilit7 that fees are positively associated wi:h demand and are in fact endogenous to demand, i.e. that they are higher whara school directors per:la.e higher demand, thus cannot be ignored. A: Least the consistantly negacive sign indicates zhat ca:er.s ;at.--us, fees and enrollment are .egatively correlated. (The bivar-a:a correlation of the fee ducay and the fee is +.12.) 19/ See, for example, Eeynenan and Loday, 1983. Tor evidence chat school quality has a positive effect on adult earnings, see 3ehran and Birdsall, 1983. 20/ Answorth, .t. al. 21/ World 3ank Educacion Sct:or Yanorandum, Anne- 11, p. :S. See also Psacharopoulos, 1977. - 31 - 22/ For scme conments on the orgrnizational capaci:7 of these villages and implications for local financing of health and vater proJects, see Ainsworth ec. al. 23/ For a scrong statement on the problem of too uch centralization in educational systems, see Schultz, 1980. - 32 - Avvendix A VARIABLE MANS AND STANDARD DEVIATONS (Variables used in Table I of tax: b/) Dependent Variables mean Standard deviacion Jhaer -f persons enrolled -1.89 2.06 mins number of persons aged 6 to 14 lumber of persons enrolled .433 .656 divided by persons aged 6 co 14 Rousehold variables Household income in '00 of Mr 8138.44 8334.11 3atural logarithm of 8.67 .986 household incoma in 'CO of IF Number of persons aged 15 to 50 2.62 1.69 divided by persons aged 6 to 14 School ouality variables a/ Average educaciin of teachers 11.2 .246 Student-teacher ratio 34.1 7.62 Index of number of grades in school 2.92 1.14 Number of books per classroos 2.15 3.47 lumbar of weeks since catchers paid 10.0 6.28 Proportion of students passed prizar7 school leaving test .327 b/ School fee variables a/ Natural logarithm of fees to paren: 2.0 .353 association in '00 MY D==y: fee charged at school .504 .502 Distance to school variables Kilometers !rom village center to 8.30 9.58 nearest primar7 school Dummy: school located ou:side .114 .319 village but within 5 kilomacers Dummy: school 6-10 kiloUeters away .106 .309 Dumy: school 11-15 kilomecers away .114 .319 D y: school 16-20 kilometers avay .0976 .298 Dumy: school more thun 20 kilometers away .146 .355 a/ Measured at village level based on closes: school. b/ Sample size is 123 excepc for proportion scudents,passing :asz variable, for which sample size is 113. Refarences Ainsworth, Martha. "The Demand for Health and Schooling in Mali: Results of the Comunity and Service Provider Survey." World Bank, Country Policy Department, draft, February 7, 1983. Ainsworth, Martha, Francois Orivel and ?unam Chuhan. "Cost Recovery for Health and Water Projects in Rural Mali: Household Ability to Pay and Organization Capacity of Villages." World Bank, Country Policy Department, April 1983. Backer, Gary S. "A Theory of the Allocation of Ti=&." Economic Journal 75 (September, 1963), 493-517. 3ahrman, Jere R. and 14ncy Birdsall. "The Economics of School Quality: Quantity Alone is Misleading," World Sank, Country Policy Departmenc Discussion Paper 83-1 (1), 1983 (American Economic Review, forthcoming.) Birdsall, Nancy. "Strategies for Analyzing Effects of User Charges .ia the Social Sectors." World Bank, Country Policy Departmar.t, era!t, July 1982. 3irdsall, Nancy and Pua. Chuhan. "a-2.ingness to Pay for Health and Wate: tl Rural 11.ali: Do WT? Questions Work?" World Sank, Country Policy Department, draft, March 1983. Ecole Nationate do Medicine at do Pharmacia du Mali.: "Evaluation Sanicaire des carmles de Kanieba, Safoulabe, Kica: Rpport Preliminaire." February 1982. Gray, Clive and Andre Martens. "The Political Economy of the 'Recurrert Cost Problem' in the ;est African Sahel." Mi=emo, January 1982. Hayneman, Stephen and William Loxley. 1983. "The Impart of Prinary School Quality on Academic Achievement Across Twenty-Nire High and Low- Income Countries". American Journal of Sociology, fofthcoming. Psacharopoulos, George, "Yeasuring the Walfare Effec:s of Educational Policies." pp. 75-94 in Public Economics Human Resources, eds. Vctor Halberscadt and Anthony J. Culyer. Paris: Ediclons Cujas, 1977. Republique du all, UNESCO. L'eiucation au Mali: Problemes, ?erspectlves ec Priorities. 3amako, May 1981. Schultz, T.W. Investing in People: The Economics of P-oulation Quality. Berkeley, California, Universi:y of California Press, 1960. - 34 -- ãorld 3anàc, Mal± Rescarci ?%oIec: (R2O 672-72) siaf!, "`h. - ali 21a.11.1oL, Comuni:y aad Scrvica Provider Surveys: Documentacion". C.try Polcy Depart:anc, 1983. .or1d 3ank, .ducacion Sector 2a=nrandu, Republic of Mali, Decamber 4, 1981. Rapor: .No. 3637-Y.I. - 35 - Martha Ainswor :, 'The Dezznd for Health and Schooling in %rli - Results of the Coamunity and Service P"ovidez S-rvey", March 1983. Martha Ainsworth, Francois Orivel and Punam Chuhan, "Cost Recovery for Health and Water Projects in Rural Mali: Household Ability to Pay and Organizational Capacity of Villages", April 1983. Nanc7 Birdsall, "Demand for Primary Schooling in Rural Mali: Should User Fees Be Increased?", April 1983. Nancy Birdsall and Punam Chuhan, -.illingness to Pay for iealth and ater in Rural Mali: Do WTP Questions Work?", March 1983. World Bank, Research Projece No. RPO 672-72, "The Mali Household, Comunity and Service Provider Surfays: Documentation", April 1983. Forthcoming: "The Determinanti of Health Status and Deand for Health Care in Rural Mali". "Household Production and the 'Returns' to Health and Schooling Inves ten ts". "Policy Issues for Schooling in 4ali".