HEALTH MAINTENANCE BY WORKERS IN RURAL AND URBAN PRODUCTION STRUCTURES 1 WPS/

Transcription

1 HEALTH MANTENANCE BY WORKERS N RURAL AND URBAN PRODUCTON STRUCTURES 1 by Germano M. Mwabu 2 and Stephen A. O Connell 3 WPS/ April 20, 2001 ABSTRACT: The paper explore ome interaction between production environment and health maintenance by worker in developing countrie. We argue that rural occupation, being le highly pecialized than occupation in urban area, offer member of rural houehold a range of productive activitie requiring different combination of health capital and other input. Urban occupation, in contrat, typically admit a much maller range of activitie. We ue a highly tylized model of a worker allocation of labor time to demontrate that the non-pecialized production environment of the rural worker raie the opportunity cot of health care at low level of health, and thu weaken incentive for curative health maintenance. Health policy implication of thi reult in the context of developing countrie are drawn. 1 Work on thi project tarted in 1986, when O Connell wa a viiting Reearch Aociate at the Department of Economic, Univerity of Nairobi. He thank the Economic Department for it generou hopitality and the Univerity of Pennylvania for leave upport. We thank Jere Behrman for comment on the original draft. An earlier verion of the paper wa completed in Augut 1989 and revied in April Mwabu thank the nternational Health Policy Program for financial upport during the period he worked on the project. 2 Aociate Profeor, Department of Economic, Univerity of Nairobi. Mwabu@formnet.com. 3 Profeor, Department of Economic, Swarthmore College, and Reearch Aociate, Centre for Study of African Economie, Univerity of Oxford. teve_oconnell@warthmore.edu.

2 1. ntroduction The purpoe of thi paper i to explore ome interaction between production environment and health maintenance by worker in developing countrie. We tart with two obervation. Firt, health tatu in developing countrie i generally lower in rural area than in urban area (Ribero (1999), Anker and Knowle (1980), Koinange (1982), Bigten (1977), Republic of Kenya (1986)). Second, a key characteritic differentiating the urban from the rural work environment, particularly in developing countrie, i the lower degree of pecialization of labor in the rural ector. We argue that rural occupation, being le highly pecialized, offer member of rural houehold a range of productive activitie requiring different combination of health capital and other input. Urban occupation, in contrat, typically admit a much maller range of activitie, and thee activitie themelve offer limited poibilitie of ubtitution between health capital and other input. The limited range of productive opportunitie for an individual in a pecialized work environment uch a an urban area, reduce the opportunity cot of labor time. A mechanic or an electrician for example, ha fewer productive activitie outide hi pecialized trade relative to a rural trader, who for intance may engage in a variety of commercial and agricultural activitie. We how that in the particular cae of ickne (low level of health) when both worker are not in their uual employment, the reervation wage for the urban worker i lower than that for the general worker in a rural etting, and explore formally the health maintenance conequence of thi wage differential. Since health care typically require an allocation of time (e.g., a hopital viit), the maintenance of health hould be affected by the worker perception of the opportunity cot of time away from production. The role of time a a device for rationing health maintenance ervice in developing countrie ha been well documented in the literature (ee, e.g., Acton (1973, 1975), Akin et al. (1986), Dor et al. (1987), Mwabu (1989), Alderman and Gertler (1989) Ribero (1999)). n Nigeria, for example, 50 percent of a ample of working women cited the time contraint a the major reaon for not getting their children immunized (Akin, et al. (1985)). Since human capital theory would ugget that the poor have a lower opportunity cot of time than the non-poor, a major policy implication of thee finding ha been that rationing of health ervice through queue, or other mechanim involving ubtantial time expenditure, benefit the poor more than the non-poor. We will argue, intead, that the production environment of rural houehold raie the opportunity cot of a ick worker time above the level that the ame worker would confront in an urban production tructure. The rural worker thu ha a weaker incentive for curative health maintenance than hi or her urban counterpart. The relatively high opportunity cot of health care in rural area come from the fact that the rural worker whoe health ha deteriorated can generally find ome productive ue for hi impaired health capital. For example, a rural worker uffering from a mild attack of malaria can hift from harveting to lighter work uch a caring crop predator. n contrat, the opportunity cot of health maintenance for urban worker decline more rapidly a health deteriorate, ince other input cannot be ubtituted for health capital, epecially in the hort run, without a ubtantial lo in income. 4 While thi i le true for the urban informal ector than the formal ector, 4 Thi effect i trengthened by the availability of ick leave for urban worker and the greater prevalence of health inurance in urban area. The effect of thee additional intitution on health maintenance ha been tudied in the literature; ee, e.g., Feldtein (1973), Arrow (1963). Peron 1

3 we believe that our ditinction till hold when one compare the urban informal ector with the rural ector. The ability of the rural worker to chooe among a range of activitie ha three baic ource. Firt, the rural worker ha ome latitude over the equencing of production activitie, even within the contraint impoed by eaonal requirement (Chamber, et al. 1979). For example, a farmer with failing health can undertake relatively heavy farm activitie in the morning uch a plowing, and le demanding activitie in the afternoon, uch a repair of farm tool. Second, and more importantly, production and health maintenance deciion are made at the level of the houehold, where the low level of pecialization of labor permit the exchanging of tak among family member (Mwabu and Wang ombe (1987), Ribero (1999)). Thi i true even though economic force or ocial cutom may impoe contraint on the intra-family allocation of tak. Contraint of thi type, uch a male/female ditinction in nonmarket activitie and the reerving of market activitie for adult, typically rule out only a ubet of the poible exchange of tak among family member. 5 The third ource of the ability of a rural worker to chooe from among a variety of occupational activitie i the informal nature of hi work environment, which unlike that of hi urban counterpart, i flexible, in the ene that it i not rigidly regulated by employment contract or norm (ee Ribero, 1999). n the next ection, we ue a imple analytical model to bring out the key relationhip between production tructure, health maintenance, and health tatu. Section 3 dicue empirical implication of the model and Section 4 conclude the paper. 2. The model n tandard neoclaical labor economic, worker chooe between two activitie, wage-earning labor and leiure. The optimum allocation of time i to work until the marginal utility of conuming the wage good equal the marginal utility of leiure. To tudy the interaction between production environment and health maintenance, we begin by making four obervation. Firt, certain occupation, uch a that of the rural peaant worker/producer, offer the individual a range of ditinct productive activitie with different input-output relationhip. Second, the activitie worker chooe among (including leiure) produce not only pecuniary reward but alo outcome in term of health tatu. Third, worker can chooe to allocate time not only to productive activitie or leiure, but alo to health maintenance, an activity we will view a ditinct from leiure. 6 with a job that provide ocial ecurity may be more likely to take day diabled than individual who are elf-employed and uncovered by ocial ecurity, for whom it may be more cotly not to attend their job. (Ribero (1999, page 4). 5 For example, adult and children can exchange certain non-market activitie, and male and female adult can exchange certain market activitie. 6 n reality, the boundary between leiure and health maintenance i not very clear. The neoclaical utility theory doe not require a decription of the leiure activity beyond whatever i implicit in the condition that the utility function be a trictly quai-concave function of conumption and leiure. Operationally, the theory identifie leiure a anything that i not work. Thi concept of leiure clearly include our narrowly defined health maintenance activity (ee below), along with a number of other activitie like leeping and eating that we will continue to claify implicitly a leiure, even though they may be more properly thought of a health maintenance. 2

4 Finally, effort can be endogenou variable: there are many activitie in which the worker i free to alter the productivity of hi time by changing the intenity of work. Each of thee obervation point to a pecific modification of the tandard conumer theory. Our baic idea, however, can be made precie in a imple model in which we abtract from the leiure and effort choice and focu on the allocation of labor time between productive activitie and health maintenance. We therefore analyze uch a model here, recognizing that empirical work will require a richer model. Activitie, occupation, and health tolerance We define an occupation uch a that of a contruction worker or peaant farmer, a a ubet of the et of all poible productive activitie. An activity i a way of combining labor and other available input to produce (1) a pecuniary value and (2) a change in the worker health tatu. The pecuniary value produced by an activity i the direct payoff received by the worker, which might be a fixed hourly wage for a contruction worker or the value of a mended plough or plowed land for a peaant farmer. For implicity, we aume that each activity require exactly one unit of labor time. The worker health tatu i ummarized by a ingle meaure,, that i a ufficient tatitic for the health capital that the worker can bring to bear in any activity. Thi i clearly a implification. Without further lo of generality, we take to lie in the interval S = [0,1]. Throughout the enuing analyi, the parameter, whoe value are retricted to lie within a unit interval, repreent the health tatu of worker. Activity i in period t, then, i defined by a payoff function A ti : S SxR that i a mapping from the worker health tatu to the joint product of a pecuniary reward r( t ;i) R and a new health tatu t+1 ( t ;i) S. The mapping A ti are non-tochatic, i.e., there i no uncertainty about the pecuniary and health effect of each activity. 7 We aume that both r and t+1 are non-decreaing function of t. An important characteritic of productive activitie i what we will call their health tolerance, i.e., the extent to which the worker productivity in the activity i reilient to the reduction in work intenity that accompanie a decline in health tatu. Formally, we define health tolerance aociated with a given activity a follow: Definition Activity i ha higher health tolerance than activity k if the following propertie hold: (i) (ii) (iii) r(;i) r(;k) for ome S and r( ;i) r( ;k) for ome S f r(;i) r(;k) for any S, then r( ;i) r( ;k) for all [0,]. t+1 ( t ;i) t+1 ( t ;k) for all t S. Property (i) i the requirement that neither activity trictly dominate the other in term of pecuniary payoff at all level of health. n other word, while activity k i at leat a 7 f worker were rik-neutral, A ti could imply refer to expected value. n thi cae our analyi could be trivially extended to incorporate uncertainty. To incorporate rik avere behavior, which play an important role in a variety of theorie of labor upply, from implicit contract theory (Azariadi (1975)) to theorie of ubitence production mode (Berry (1977)), we would need a more ubtantial modification of the analyi. 3

5 productive a activity i at ome health level(), there mut be ome level of health at which the individual can do jut a well in direct pecuniary term by allocating hi labor to activity i. Property (ii) tate that if activity i i more productive than activity k at any level of health, it remain at leat a productive a health deteriorate further. Together, propertie (i) and (ii) imply that the pecuniary reward function r(;i) and r(;k) can cro at mot once for S, with r(;i) cutting r(;k) from above. 8 Figure 1(a) 1(c) give example of activitie atifying the ingle croing property. Figure 1(d) give a cae where thi property i not atified. Property (iii) give the econd requirement for activity i to have higher health tolerance than activity k: the productivity advantage at low level of health mut not be achieved at a cot in term of future health tatu. For the remainder of the dicuion, we will make the implifying aumption that t+1 ( t ;i) = t for all productive activitie. Under thi aumption we can conclude that activity 1 ha higher tolerance than activity 2 in Figure 1(a) 1(c); in 1(d), the two activitie cannot be ranked. Notice that health tolerance i tranitive in the ene that if activity i ha higher tolerance than activity j, and j ha higher tolerance than activity k, then i ha higher health tolerance than k. The health tolerance relation doe not, however, provide a complete ordering to the et of activitie, ince not all pair of activitie can be ranked. We therefore cannot repreent health tolerance by an ordinal quantity in the ame way a the preference relation can be repreented by a utility function in conumer theory. Along with hi occupational et of production activitie, which we denote Z j = {i: A i i an activity in occupation j}, each worker ha acce to a health maintenance activity. While health maintenance i in reality a complex proce involving a range of alternative activitie (epecially in developing countrie; ee Mwabu (1986)), we will implify matter coniderably by characterizing health maintenance a a ingle activity, identical for all occupation. Thi activity require an input of labor time (e.g., a hopital viit) and produce a pecuniary reward that i typically non-poitive (the uninured portion of the hopital fee) and an improvement in health tatu. 2.2 The health maintenance deciion The worker problem, then, i to chooe the activity to which he will allocate hi current unit of labor time, given hi current health tatu. The problem i inherently intertemporal, given the preence of health capital. n fact, the labor upply deciion can be viewed a the olution to an invetment problem, ince any activity involve a trading off of current and future payoff through modification in the tock of health capital. We emphaize the analogy with invetment by auming that the worker receive no direct utility from hi tock of health capital. The payoff to health maintenance i therefore the increae in future pecuniary reward due to a higher tock of health capital. The intertemporal apect of the labor/health maintenance deciion can be captured by auming that the worker i endowed with an initial health tatu 1, and two period of labor time. The problem i to chooe activity in each period o a to 8 The reward function mut touch at leat once becaue of property (i). Since the inequalitie in (ii) are weak, they do not have to cro; they can coincide for all above ome level. 4

6 maximize lifetime utility. For implicity, we aume that lifetime utility i a linear function of lifetime pecuniary payoff. 9 Letting V tj ( t ) be the maximized value of remaining lifetime payoff for occupation j under an optimal choice of activity in time t, the lifetime payoff in period 1 atifie (1) V 1j ( 1 ) = Max r( 1 ;i) + V 2j ( 2 ( 1 ;i)), {i Z 1j } where r( 1 ;i) R and 2 ( 1 ;) S are the pecuniary and health output, repectively, of activity i if initial health i 1. The econd period value function, V 2j, i given by (2) V 2j ( 2 ) = Max r( 2 ;i). {i Z 2j } The olution to the worker optimization problem depend on the tructure of the equence {Z 1j, Z 2j } of available productive activitie. Our tak i therefore to characterize the activity et aociated with urban and rural occupation in developing countrie. n urban occupation, worker chooe from the ame mall et of activitie each period, o that Z 1U = Z 2U. The rural occupation differ in three key repect. Firt, it contain a larger menu of activitie. Thi i what we mean by the rural work environment being le highly pecialized. Second, the rural worker can ubtitute acro activitie with different health tolerance. We incorporate thi poibility by auming that the rural occupation contain at leat one activity with higher health tolerance than any of the urban activitie. Third, the equence of activity et for the rural occupation may contain eaonal retriction that a range of different activitie be completed over the planning horizon. A we will ee below, thi lat feature complicate but doe not change the concluion. A final technical aumption we require i that the maximum pecuniary reward attainable in a ingle period in the rural occupation, Max {r(1;i): i Z tr }, i not greater than the maximum reward attainable in the urban occupation. Given that the average return to labor i typically higher in urban than rural area in developing countrie (conditional on employment for the urban worker; ee, for example, Harri and Todaro (1970) and the ubequent literature), thi aumption i conitent with our econd claim, which implie that rural worker with low level of health capital are more productive than urban worker with the ame low health tatu Rural/urban health differential Given thee aumption, we can now tate our baic reult and then illutrate and interpret it in the context of a imple example. The quetion we addre i the following: if initial health level 1 of urban and rural worker are independently and identically ditributed on S, and the health maintenance activity i identical for the 9 ncorporating a direct payoff to health capital (beyond it marginal product in production activitie) would not change the analyi qualitatively. See Groman (1972) for a dicuion of the invetment and conumption component of the health maintenance activity. One could alo eaily add poitive time preference, non-zero real interet rate, and/or curvature of the utility function without qualitatively changing the reult. 10 The aumption doe not follow trivially, however, ince the average health level i alo higher in urban area. 5

7 two occupation, which occupation will have the higher average health level in period 2? n other word, what i the effect of production tructure on health maintenance? Propoition f initial health level are independently drawn for all worker from an underlying ditribution f( 1 ) and the health maintenance activity i identical for all occupation, there i ome number C (not necearily poitive) uch that if the health maintenance fee exceed C, the average health level of rural worker will be lower than that of urban worker in period 2. For illutrative purpoe, we tudy here an example in which Z tu contain a ingle activity and Z tr two activitie. For implicity, we aume that both occupation hare a common activity A 2, and that the rural occupation ha a econd activity A 1 with higher health tolerance than A Uing the et of activitie Z tj, we can define the ingle period pecuniary production function for occupation j, g j (), a the maximum pecuniary reward in the current period for a worker with health tatu. Thi function, plotted in Figure 2(a) and (b), i imply the upper envelope of the r(;i) for each occupation. The function g j () ha a natural interpretation: ince g j () i the direct pecuniary reward given up by pending a unit of labor time in health maintenance, g j () + C i the opportunity cot of health maintenance in the current period, where C i the fixed health maintenance fee. The optimal choice of activity in period 1 for a worker with health tatu i now eay to etablih. n period 2, there i no health invetment component to the labor upply deciion, o the worker imply chooe from the et Z 2j the production activity yielding the pecuniary reward g j ( 2 ). We therefore have V 2j ( 2 ) = g j ( 2 ), and the problem in period 1 (ee equation (1)) i to maximize (3) r( 1 ;i) + g j ( 2 ( 1 ;i)) by chooing i from the et {1,2,3} where i = 3 i the health maintenance activity. The problem come down to one of comparing lifetime utility under the optimum production choice in period 1, V P 1 j ( 1 ), with lifetime utility from chooing health maintenance in period 1, V H 1 j ( 1 ). Since the optimum productive activity in period 1 yield g j ( 1 ), thee function are given by P (4) V ) = g ( ) + g ( ) = 2 g ( ) 1 j ( 1 j 1 j 1 j 1 H (5) V ( 1 ) = C + g (1), 1 j j for j = U, R. Equation (4) and (5) incorporate our aumption that productive activity leave health unchanged ( 2 = 1 if production i choen) and that health maintenance involve paying a fee of C to raie health tatu to 2 = 1. P H There may be level of health capital at which V1 j ( 1) = V1 j ( 1), o that the worker i jut indifferent between production and health maintenance in the current period. f health capital ha trictly poitive pecuniary marginal product (i.e., g j () i monotonically increaing), there can be at mot one uch point of indifference for 11 Recall that thi mean that productivity i higher in A 1 than in A 2 at low health level. 6

8 (0,1). We denote thi level of health capital a 1 j. From (4) and (5), it i apparent that 1 j i the olution to 1 (6) g j ( 1 j ) = ( g j (1) C), j = U, R. 2 Figure 3 illutrate the determination of 1 j. The right-hand ide of equation (6) i reflected onto the vertical axi tarting with the value of g j (1) and uing the line in the econd quadrant. The cutoff value 1 R and 1 U can then be read directly from the function g j ( 1 ). Given that < 1 U, it i traightforward to how that rural worker will have a lower average health level in period 2 than urban worker. Since the empirical ditribution of urban and rural worker by initial health tatu converge to the underlying ditribution f( 1 ), the average health tatu of occupation j in period 2 converge in probability to the true expectation E( 2j ), which i given by (7) E( 2 j ) = E( + E( + E( 2 j 2 j 2 j 0 < 1U 1 j < < < 1 j 1 j < ) Prob(0 < 1U ) Prob( < 1) Prob( 1U 1j < < 1j 1j ) 1U 1U ). ) Since urban and rural worker with health tatu on the interval ( 0, ) and ( 1 U,1) make the ame health maintenance deciion, the only difference between E( 2U ) and E( 2R ) come from the fact that urban worker on ( 1, R 1U ) chooe health maintenance, while their rural counterpart, whoe opportunity cot are higher, chooe production. t follow that (8) E < < ) < E( < < ), ( 2R 1U 2U 1U 1U and therefore that E( 2R ) < E( 2U ). n Figure 3a, the upper envelope of rural activitie (g R ) meet the envelope of urban activitie (g U ) from above and never goe below it. Uing our definition of health tolerance, it follow that the rural occupation a a whole exhibit higher health tolerance, in thi example, than the urban occupation. The exitence of uch a ranking turn out to be a ufficient but not a neceary condition for lower health maintenance by rural worker. To ee thi, rewrite equation (6): (9) gi ( 1) C 2 = 1. g (1) g (1) i i The ick worker therefore compare normalized lifetime income when ick the lefthand ide with normalized lifetime income net of health maintenance cot (the right-hand ide). Health maintenance i worthwhile only if the right-hand ide exceed the left-hand ide. Now uppoe that the rural occupation ha higher health tolerance than the urban one. With greater health tolerance, the fact that average rural income 7

9 are below average urban income implie that g R (1) g U (1); otherwie rural productivity would dominate urban productivity at all health level. t follow from thi that (a) for ufficiently low health level, foregone productivity g i ()/g i (1) i maller for the rural worker than the urban worker; and (b) that for any fee C, the burden of the fee, C/g(1), i at leat a great for the rural worker a for the urban worker. Thee obervation reinforce each other to produce R < U in equation (9), providing the critical tep in our propoition. A health tolerance ranking i not, however, neceary for our reult. A pair of ufficient condition for the equality in equation (9) take place at a lower level of in rural area than in urban area i (1) that average income be higher in urban than in rural area and (2) that normalized productivity g i ()/g i (1) (rather than abolute productivity) atify the ingle-croing property. The latter requirement imply tate that there mut be ome level of low health below which the rural worker retain a greater hare of hi productivity when healthy than doe hi urban counterpart. Thi doe not require that the rural worker be abolutely more productive than the urban worker, even at low health level. 2.4 Generalizing the production tructure n the cae where Z Z 2R, the analyi proceed along the ame line. Figure 4 indicate why we get imilar concluion a to relative health maintenance. The figure how a variety of alternative equence of pecuniary reward function {g 1j (), g 2j ()}. The top panel how the urban occupation, with g 1U = g 2U. Panel (b) how the cae we have jut analyzed, where the rural occupation ha more activitie but g = g 2R. Panel (c) and (d) give the two poible ordering of activitie in the rural occupation. There are two alternative to the aumption that Z = Z 2R : the rural occupation could be trongly eaonal, a in (c) or (d), or it could allow flexibility a to ordering of tak but require that both tak be completed over the planning horizon. The econd of thee alternative would allow the worker to chooe between the ordering (c) and (d) but would rule out performing the ame tak twice. Conider firt the eaonal alternative, and compare the rural worker incentive for health maintenance with that of the urban worker. f the low health tolerance activity mut be completed in the current period (panel c), then clearly the rural worker will chooe the ame or lower level of health maintenance than hi equally healthy urban counterpart. Thi i becaue the two have the ame current opportunity cot, while the urban worker ha a higher future payoff to health improvement. The ame reult of lower health maintenance emerge in the off-eaon (panel d), although in thi cae the key factor i not future payoff to health maintenance which are identical for the two occupation but current opportunity cot. The rural worker ha more productive ue for low level of health capital and i thu le likely to eek medical care if health i poor. 12 Since the rural worker incentive for health maintenance i lower under either eaonal pattern, it follow that health maintenance will be lower if the production tructure i uch a to allow the worker to chooe the ordering of activitie. n thi cae, a rural worker with health tatu, compare either panel (c) or (d) (the two 12 For 1 > C, where C i the health tatu at which r 1 () and r 2 () cro, thi tendency i revered: the urban occupation ha a higher current opportunity cot. The reult we can etablih for panel (d) i therefore imilar to propoition 1, i.e., rural worker will have lower health maintenance level if C i ufficiently high. 8

10 ordering yield identical lifetime payoff, conditional on producing in period 1) with the alternative of eeking health maintenance and then performing the low health tolerance, high payoff activity in period 2. For 1 < C, lifetime payoff producing in period 1 i higher than the urban worker, and therefore the incentive for health maintenance i lower. 2.6 Extending the model We give a brief dicuion here of two poible extenion of our analyi. The firt i to model the migration deciion of worker, taking urban and rural production tructure a given. The econd i to allow the production tructure themelve to be jointly endogenou. What hould we expect to occur if urban and rural worker have the option of migrating between ector? Thi i a complicated iue, and we need to make ome implification. One way to proceed would be to have the migration deciion occur before the firt period of work. Abtracting from migration cot, lifetime utility for a worker with initial health 1 would then be (10) V ) = Max [ V ( ), V ( )], 1( 1 1 1U 1 where V 1j ( 1 ), the maximized utility of a worker chooing to locate in ector j, i given by P H (11) V ) Max [ V ( ), V ( )]. 1 j ( 1 = 1 j 1 1 j 1 The heterogeneity of worker here make it difficult to formulate a ueful model of locational equilibrium. Thi i a problem that occur in the literature on education and human capital a well; by analogy to Mincer (1974) analyi of chooling, we might aume that there are a large number of worker of each health tatu, and that migration i the proce by which worker arbitrage the return to health capital in each ector into equality. Equilibrium would then be characterized by (12) V ) V ( ), ( 1 = 1U 1 for all level of health capital oberved in both ector. n order to olve equation (12) for the equilibrium ditribution of worker of type 1, we would have to relate the payoff function r(;i) to the ditribution of worker by health tatu in each location. The Harri-Todaro (1970) model ugget a poible approach; there, the effect of migration on ectoral real wage come from diminihing marginal product of labor. We might think of the r(;i) in the urban ector (for example) a parameterized by the total amount of effective labor in that ector; if f U () i the number of urban worker of health tatu, total effective urban labor i approximately (13) L U = f U ( ) d E

11 Diminihing marginal product of labor would be repreented by making r(;i) an U increaing, concave function of L E. The problem would then be to olve for the equilibrium ditribution f U (). A econd poible extenion of the analyi would be to allow the ditribution of health capital and urban and rural production tructure to be jointly endogenou. Hitorically, the proce of economic development ha often involved a imultaneou accumulation of capital and ectoral hift out of rural and non-market activitie toward urban, market activitie. We make two obervation. Firt, accumulation of health capital and human capital i an important component of the overall capital accumulation proce that take place during development. Second, the ectoral hift that occur in the development proce i not imply a migration of reource from one exiting production tructure to another; rather, it involve the introduction of new activitie in the expanding ector. ndutrialization involve not jut hifting worker out of agriculture but alo development of an urban production tructure with higher degree of labor pecialization, more elaborate work rule, and a greater preponderance of market activitie than in the rural tructure. Development of thee activitie may itelf be facilitated by accumulation of human capital in the form of health and education. Thi ugget a model of the development proce a the movement from a one-ector economy, with the introduction and growth of an urban productive ector jointly determined with the overall capital accumulation proce. 3. Empirical iue There are everal approache to teting the above model. One method require an empirical demontration of the idea that activitie in rural area are more tolerant to poor health than urban-baed activitie. f thi propoition i true, for a given activity (that i common to rural and urban area), the health tatu of the worker engaged in that activity would rie a one move from rural to urban area. We have hown that thi reult may not merely be a reflection of locational deciion of worker: at low level of health, urban worker invet more in better health becaue they face lower time cot of health maintenance. Teting the effect of production tructure on health maintenance i complicated by the need to control for effect of migration and eaonality. The econd, and more direct approach to teting the baic propoition of our model i to etimate the effect of the number of activitie performed by worker in a given period on health tatu. The number of activitie performed i a meaure of a worker degree of pecialization. More pecialized worker will have a maller bundle of activitie than non-pecialized worker. n our model, a worker health tatu improve a the number of activitie performed decreae, i.e., with occupational pecialization. Panel data on worker occupation and rural-urban migration are required to tet thi propoition. A final teting method involve meauring the effect of a deterioration in health tatu on earning of worker in pecialized (urban) and non-pecialized (rural) activitie. The baic propoition of the model would be rejected, if a fall in health tatu (a meaured by diability day for example) were to reduce earning in rural area by a maller magnitude than in urban area. 10

12 4. Concluion The focu of thi paper ha been on interaction between health tatu, health maintenance and production tructure. The main reult of the paper i that production tructure affect health maintenance cot, and thu could be an important determinant of demand for health ervice. n developing countrie, difference in rural-urban production tructure could account for a ubtantial variation in health tatu in the population. The idea that at low level of health, the opportunity cot of time i higher in rural than in urban area, conflict with the tylized fact that labor income (particularly in developing countrie) i higher in urban area (ee Harri and Todaro, 1970). We have hown that oberved labor income i the average of income from activitie performed by a worker at low and high level of health. n doing o, we aumed that labor income i greater the higher the level of health. Thu, irrepective of reidence tatu, a worker mean income from activitie performed at low and high level of health, i lower than income earned at high level of health. However, becaue of gain from pecialization, the mean labor income i higher in urban than in rural area. n general, therefore, the opportunity cot of time i higher in urban than in rural area. At low level of health, the ituation could be revered. We explain thi reveral by auming that rural production tructure are more tolerant of declining health than rural production tructure. A health deteriorate, a rural worker i more likely to find an activity that i compatible with hi health tatu than i an urban worker. Hence, a rural worker face higher time cot of treating an illne compared to hi urban counterpart. The policy implication of thi finding i that in developing countrie, where the majority of the population i rural, ubidie for rural health ervice may be needed to facilitate the achievement of the international goal of good health for all (WHO, 1978). Specifically, policie that reduce the time cot of acceing and uing rural health ervice hould be implemented. Our tudy ugget that the poor in rural area face much higher time cot of health maintenance than previouly thought. 11

13 Reference Acton, J. P. (1973), Demand for Health Care When Time Price Vary More Than Money Price, Rand Corporation No. R-1189-OEO/NYC, Santa Monica, CA. Acton, J. P. (1975), Nonmonetary Factor in the Demand for Medical Service: Some Empirical Evidence, Journal of Political Economy 83 (3): Akin, J. S., C. C. Griffin, D. K. Guilkey and B. M. Popkin (1985), The Demand for Primary Health Service in the Third World (Totowa, N.J.: Rowman and Allanheld). Akin, J. S., C. C. Griffin, D. K. Guilkey and B. M. Popkin (1986), The Demand for Primary Health Care Service in the Bicol Region of the Philippine, Economic Development and Cultural Change 34(4): Alderman, H. and Gertler, P. (1989), The Subtitutability of Public and Private Health Care for the Treatment of Children in Pakitan, LSMS Working Paper No. 57, The World Bank, Wahington, D.C. Anker, R. and J. Knowle (1980), An Empirical Analyi of Mortality Differential in Kenya at the Macro and Micro Level, Economic Development and Cultural Change 29: Arrow, K.J. (1963), Uncertainty and Welfare Economic of Medical Care, American Economic Review 53: Azariadi, C. (1975), mplicit Contract and Underemployed Equilibria, Journal of Political Economy 83: Berry, S. S. (1977), Rik and the Poor Farmer (Technical Aitance Burean, USAD). Bigten, A. (1977), Regional nequalitie in Kenya, ntitute for Development Studie, Univerity of Nairobi, DP. No. 302, Nairobi. Chamber, et.al. (1979), Seaonal Dimenion to Rural Poverty: Analyi and Practical mplication, Dicuion Paper 142, ntitute of Development Studie, Univerity of Suex, Brighton. Dor, A. P. Gertler and J. van der Gaag (1987), Non-Price Rationing and Medical Care Provider Choice in Cote d voire, Journal of Health Economic 6(4): Feldtein, M. S. (1973), The Welfare Lo of Exce Health nurance, Journal of Political Economy 81, Part : Groman, M. (1972), On the Concept of Health Capital and the Demand for Health, Journal of Political Economy 80(2):

14 Harri, J. R. and M. P. Todaro (1970), Migration, Unemployment and Development: A Two Sector Analyi, American Economic Review 60: Koinange, W. (1982), A Report by the Director of Medical Service on Health Statu in Kenya, 1979 (Minitry of Health, Nairobi). Mincer, J. (1974), Schooling, Experience and Earning (New York: National Bureau of Economic Reearch). Mwabu, G. M (1986), Health Care Deciion at the Houehold Level: Reult of a Rural Health Survey in Kenya, Social Science and Medicine 22(3): Mwabu, G. M (1989), Nonmonetary Factor in the Houehold Choice of Medical Facilitie, Economic Development and Cultural Change 37(2): Mwabu, G. M and J. K. Wang ombe (1993), Agricultural Land Ue Pattern and Malaria Condition in Kenya, Social Science and Medicine 37(9): Republic of Kenya (1986), Economic Survey (Minitry of Planning and National Development, Nairobi). Ribero, Roccio (1999), Earning Effect of Houehold nvetment in Colombia, Center Dicuion Paper No. 810, Economic Growth Center,Yale Univerity, New Haven, CT. World Health Organization (1978), The Alma Ata Conference on Primary Health Care, Geneva, Switzerland. 13

15 Figure 1 r(;i) (a) r(;i) (b) r(;1) r(;1) r(;2) r(;2) r(;i) r(;i) (c) (d) r(;1) r(;2) r(;1) r(;2)

16 Figure 2 r(;i) (a) Rural Occupation r(;i) (b) Urban Occupation b b g R () a g R ()=r(;2) r(;1) r(;2)

17 Figure 3 g R,g U 45 o g R (1)=g U (1) 1/2[g j (1)-C] lope -1/2 1U 1 1 -C/2 16

18 Figure 4 g( 1 ) g( 2 ) t=1 t= g( 1 ) g( 2 ) t=1 t= g( 1 ) g( 2 ) t=1 t= g( 1 ) g( 2 ) t=1 t=