The Economic Impact of Health Care Provision: A Preliminary CGE Assessment for the UK 1. University of Nottingham

Transcription

1 The Economc Impact of Health Care Provson: A Prelmnary CGE Assessment for the UK 1 Martne Rutten, 2 Adam Blake 3 and Geoffrey Reed 4 Unversty of Nottngham 1. Introducton Ths paper presents some prelmnary results from a CGE model of the nteractons between government-provded health care and the outputs of non-health goods and natonal welfare n a small open economy. The effects on welfare of hgher health provson come through two man channels: (a) the drect gan from ncreasng the well-beng of the populaton, and (b) the ndrect effects of an ncrease n the sze of the effectve (.e. able to work ) endowments of sklled and unsklled labour for use n non-health actvtes. A hgher level of health provson enlarges effectve labour supply n the short term by augmentng the aggregate workng tme of current workers (ncludng wastage by premature death). It does so n the longer term by reducng death rates among those young who are destned to enter the work force. However, t also ncreases the number of people who are not part of the work force (the young and the retred). 5 These are an addtonal source of demand for health servces, so reducng avalablty and/or level of treatment for the current work force and so reducng ts effectve sze. Moreover, both groups of non-workers are usually recpents of transfers from the workng populaton (e.g. state benefts for chldren, state pensons for the retred), wth the assocated dstortons. We beleve the analyss s nnovatve n two respects. Frst, there have been few appled studes of the general equlbrum effects of llness/healthness on the non-health economy. Nearly all health economcs modellng s done n a partal equlbrum settng, and as such tends to focus on the mpact of polcy on current and prospectve patents. Ratonalsatons for the partal equlbrum approach nclude the relatve smallness of the health sector and the hgh degree of specalsaton (.e. ntersectoral mmoblty) of hghly sklled staff and hgh We are grateful to Rod Falvey, Doug Nelson and Dave Whynes for ther helpful comments and suggestons. The responsblty for any remanng errors or nfelctes remans wth us. School of Economcs Chrstel DeHaan Toursm and Travel Research Insttute, Busness School School of Economcs and Leverhulme Centre for Research n Globalsaton and Economc Polcy The domnant users of health care are those under 5 years of age, those over 60, and the long-term ll (accordng to the ONS (2002) some 850 mllon workdays are lost each year through ncapacty, representng 2.3 mllon people or approxmately 3.8% of the populaton n the UK.) 1

2 tech captal. Whle these may hold true, there are other strong nteractons wth the rest of the economy and wth polcy-makng, certanly n the longer term, so that t s arguable that a general equlbrum modellng may be more approprate for the analyss of some ssues. Wth respect to the labour market, the majorty of health sector staff, even the part that s sklled, s n practce nter-sectorally moble (managers and assocated staff, laboratory techncans, ancllary workers). Thus expanson or contracton of the health sector wll mpact on other sectors. Some sectors wll expand whle others contract, dependng on the relatve factorntensty of the health sector. Smlar comments apply to sectors producng ntermedate nputs for the health sector (pharmaceutcals, hgh-tech equpment). Second, whle there s a strong lterature on endogenous labour supply models (e.g. Martn, 1976, Martn and Neary, 1980), these have n the man been based on drect labour supply response to hgher wages. In ths model, changes n the effectve supply of sklled and unsklled labour come from changes n the sze of health provson (whch n ths UK-centred example s determned by the government). Whether these changes n factor supples ncrease some sectoral outputs and decrease others, or ncrease all outputs depends on whether the scale effect of ncreased factor supples outwegh the factor-bas effect of changes n the rato of sklled to unsklled labour. Usng a CGE model we fnd that an ncrease n publc health expendtures ncreases the health of sklled and unsklled labour whch, once allowng for postve health effects on effectve labour supply, enhances welfare up to 10 bllon pounds (compared to a deteroraton of 0.6 bllon pounds when health effects are absent). A surprsngly low degree of effectveness of health care n ncreasng labour market partcpaton rates suffces to ensure a welfare gan. Wth health effects absent, publc health care and sectors producng ntermedate nputs (pharmaceutcal ndustry and medcal, precson and optcal nstruments) expand, whereas prvate health care and other non-health sectors contract (prvate health care faces hgher nput costs). Sklled labour benefts from nter-sectoral reallocatons whereas unsklled labour and captal remuneratons fall. The ntroducton of ncreasngly strong postve health effects eventually leads to ncreases n the producton of all sectors, mplyng that the scale effect domnates factor bas effects. As labour becomes relatvely less scarce wages fall and rents rse. We also smulate a rse n the prce of pharmaceutcals under a scenaro n whch the government mantans a fxed health budget such that publc health care treatment levels 2

3 must fall and a scenaro n whch the government allows ts health budget to grow wth the am of mantanng prevous treatment levels under the NHS. In the frst scenaro levels of publc and prvate health care provsonng fall due to a rse n the costs of nputs. Health deterorates and allowng labour supply changes condtoned on health labour market partcpaton falls. Overall welfare falls up to bllon pounds. When the government ncreases publc health care expendtures so as to mantan prevous treatment levels, negatve health and welfare effects are mtgated; and welfare falls up to bllon pounds. Dstrbutonal effects of the counterfactual smulatons are unequal across labour categores and household types. Sklled labour at all tmes s worse off n terms of employment relatve to unsklled labour as some of the sklled consume publc and prvate health care, the latter becomng more costly and less avalable. Households wth workng members gan from hgher levels of health provson through ncreased partcpaton n the labour market, whereas non-workng households, who are solely dependent on benefts from the state, lose. An mportant lesson learnt from all model experments s that results dffer n the short run,.e. when health effects are absent, and the long run, when health effects materalse. Increasng or mantanng levels of NHS care n the short term mples crowdng out savngs and nvestments or foregong other expendtures, n ths model transfers to households, and hence leads to welfare losses. In the long term, when postve health effects emerge, such polces pay off and may lead to substantal welfare gans dependng on the effectveness of health care n mprovng effectve labour supply. They do so by ncreasng the labour market partcpaton of the workng populaton and by enhancng the tax earnng ablty of government, whch beneft both workng households (n terms of wage ncome) and nonworkng households (n terms of transfer ncome). The rest of the paper s organsed as follows. Secton 2 presents a lterature revew of CGE models appled to health care. Secton 3 uses standard dagrammatc analyss for a low dmenson general equlbrum model to provde some nsght nto the effects of changes n health provson on employment and producton n non-health sectors. Secton 4 gves a bref overvew of government provson of and polcy towards health care n the UK. Secton 5 dscusses the data used n the calbraton of the CGE model and the structure of the model tself, and secton 6 presents and dscusses the results of some counterfactuals. Fnally, secton 7 offers some conclusons and dscusses planned future work. 3

4 2. Revew of CGE models of changes n health and health provson Appled lterature focusng on general equlbrum effects of changes n health and health care on the economy s small but dverse n terms of applcaton area. The earlest type of models that acknowledge the economy-wde effects of mproved health, Basc Needs models, 6 were desgned to mplement the basc needs approach to development of the 1970s nto a comprehensve framework, wth ts overarchng goal of basc needs satsfacton. Health and health polcy fulfl only a mnor role and t has proven vrtually mpossble to dsentangle the effect of mproved health wthn counterfactual smulatons. Furthermore, Basc Needs models typcally are recursve dynamc, appled to developng countres and by vrtue of the latter, suffer from lack of data, a rather ad hoc approach to modellng of economc behavour and abstracton from several general equlbrum elements (such as endogenous prces and government budget). Externalty models account for the presence of external effects, such as health, educaton and envronmental effects, n a CGE framework. To our knowledge only one CGE model of health externaltes exsts, that by Savard and Adjov (1997). 7 Health mprovements appear n the form of mproved labour productvty by mplementng labour-augmentng technologcal progress n producton (as a functon of government expendtures on health relatve to the base year) whch nfluences the optmal combnaton of nputs n producton and relatve wages. The man am of the model, and ndeed of most externalty models, s to verfy whether the standard CGE result of (small) economc benefts from trade lberalsaton holds n the presence of postve health and educaton externaltes. The concluson s negatve as cuts n government expendture on health and educaton, amed at mantanng the government defct, have negatve spll-over effects on domestc product and publc sector employment, household ncome and welfare. In contrast to Basc Needs models, ths model s frmly grounded n mcroeconomc optmsaton behavour and accounts for varous ntersectoral lnkages, however, t too s appled to developng country ssues n whch health s only of secondary mportance. Further caveats are a lack of dynamc effects, no dstncton 6 Vanen and Waardenburg (1975) focus entrely on health care (n Tanzana) and model the workng tme effect of mproved health by postulatng that the number of people recoverng or dyng s a functon of the number of treatments, next to those who recover spontaneously. In van der Hoeven s (1987, 1988) Kenyan model health s restrcted prmarly to affect demographc varables. Kouwenaar s (1987) model for Ecuador also ncludes a labour productvty effect va labour augmentng technologcal progress. 7 A selecton of envronmental CGE models featurng sde effects on health care are: Vennemo (1997), Beghn et al. (1999), Bruvoll et al. (1999), Garbacco et al. (2000) and L (2002). 4

5 between workng and non-workng or age groups, and absence of endogenous labour supply effects (.e. the mpact of better health on workng tme). The most recent class of models of HIV/AIDS 8 assess the economc mpact of HIV (Human Immunodefcency Vrus) and AIDS (Acqured Immune Defcency Syndrome) usng (recursve) dynamc CGE analyss. Generally, ths lterature models the negatve health consequences of the pandemc by mposng exogenous demographc and behavoural scenaro s on the economy. Typcal features of the pandemc are that t reduces labour supply by skll type, factor productvty, and ncreases household and government expendtures on health care at a cost of expendtures on other goods and savngs. Under these assumptons the lterature s man fndng s that the slow-down n physcal captal accumulaton (due to lower savngs and nvestments), productvty growth, populaton growth and human captal accumulaton (due to a fall n supply and demand for educaton) reduces economc growth and results n a fall n per capta ncome n the long term compared to a fctonal No-AIDS scenaro. Relatve to Basc Needs and Externalty models, HIV/AIDS models are relatvely sophstcated n the sense that they model the varous channels through whch changes n health, albet negatve, affect the economy n greatest detal. Nevertheless, and most lkely due to the ncurable nature of the dsease, the HIV/AIDS studes abstract from any postve feedback effect from health (and other) expendtures to populaton health and labour supply. A related strand of Health Sector models 9 clams to be of the general equlbrum type, but snce the model doman spans health care markets only and abstracts from the rest of the world they are truly partal n nature. These models are typcally appled to developed countres and feature the behavour of patents, general practtoners, medcal specalsts, pharmacsts, drug producers (brand name and generc), parallel mporters, nsurance companes and hosptals and the varous nterrelatonshps between them. Specal attenton s devoted to the presence of market falures such as nformaton asymmetres between patents, 8 9 Kambou, Devarajan and Over (1992) mplement the mpact of HIV/AIDS on the Cameroonan economy by reducng growth rates for labour supply by skll type. Arndt and Lews s (2000, 2001) South Afrcan model ncorporates a varety of demographc and behavoural effects (household and government responses to pallate the negatve consequences of the pandemc). The mpact of the HIV/AIDS pandemc on the rate of skll accumulaton va reductons n educaton spendng s assessed by Arndt and Wobst s (2002) Tanzanan model and Arndt s (2003) model for Mozambque. Chatterj and Paelnck (1991) develop a purely theoretcal general equlbrum model. Canton and Westerhout (1999a, b) and Folmer et al. (1997) construct models appled to the Dutch pharmaceutcal and health care market respectvely, whch are employed to analyse fnancal reform measures. 5

6 physcans and pharmacsts (prncpal-agency problems) and mperfect competton n the market for pharmaceutcals caused by patentng. Although the detaled level of analyss of medcal care represents a constructve addton to the prevously dscussed CGE studes, ther partal character precludes general equlbrum analyses such as resource clams of health care (.e. competton for scarce factors of producton such as captal and labour), government budget mplcatons and the mpact on effectve labour supply of mproved health, whch are crucal for our understandng of the economc mpact of health provson. From the foregong analyss we conclude that, although the labour productvty effect of mproved health s well-establshed, none of the revewed models (Basc Needs models, Externalty models, HIV/AIDS models and Health Sector models) have assessed the mpact of changes n health provson on the sze and composton of the populaton, the labour force and ts mpact on producton, ncome and welfare over tme. In ths paper we attempt to model the endogenous labour supply,.e. workng tme, effect of changes n the quantty of health care provded, whle recognsng the resource clams made by the health sector n terms of captal and, more mportantly, labour nputs. 3. Effectve labour endowments and the health sector: some low-dmenson analytcs Consder a small open Heckscher-Ohln economy, endowed wth two types of labour, sklled (S) and unsklled (U) both subject to llness at gven rates. There are four sectors ( uses for factors): X and Y are conventonal tradables, H s the non-tradable health sector treatng the ll (modelled as addng value to the ll), W s the sector recordng those that become ll and are not (successfully) treated by the health sector (the watng lst ) and are so unable to work. We assume that health care s provded by the government and that ts expendture s determned poltcally (and so exogenously to ths model). The exogenous product prces determne the factor prces and hence sklled-unsklled ratos n the three producton sectors. Wthn the perod concerned, numbers and U of sklled and unsklled workers become ll (unable to work). However, the health servce successfully treats all but S and U of these (the loss of workng tme for those successfully treated s taken, for smplcty of exposton, as neglgble). Accountng for factor use (parallelng the full employment condtons for conventonal models) gves SI I W W SX + SY + SH + SW =S UX + UY + UH + UW = U (1a) (1b) 6

7 We are nterested n the effectve labour forces, S E andu E, where S = S S (2a) E E W U = U U (2b) W and S and U are the numbers of potental workers that reman unable to work. 10 W W Fgure 1 shows one possble ntal equlbrum. It s drawn on the assumptons that (a) the health sector s the most skll-ntensve sector, and the Y sector the least skll-ntensve, that the ncdence of llness s the same for both groups of workers, 11 and that the health sector allocates ts output of health treatment n proporton to the numbers of each labour type becomng ll. Fgure 1: An ntal equlbrum The maxmum possble endowments of sklled and unsklled labour are S and U respectvely. 12 Inputs nto the health sector are measured from O H, whle those unable to work are measured from O W. The government health budget purchases S and U of labour nputs at wages w and w. At that level of health provson the numbers of potental S U workers remanng on the watng lst are S and U (and by vrtue of the prevous W W H H 10 Ths s a queston of scalng. We could equally well work n terms of the numbers of worker-hours lost. It s more convenent to dscuss the ssues n terms of workers. 11 There s evdence that the ncdence of llness s hgher n the low-ncome groups, but we gnore ths for smplcty of exposton. 12 In the sense that there s no ll health, and hence no need for health provson. 7

8 assumptons are n the same proporton as the economy s endowment rato). The nner box then gves the sklled and unsklled labour avalable to work n the two tradables sectors. Measurng nputs nto the X sector from the north-east corner of the health box and nputs nto the Y sector from the south-west corner of the watng lst box allows us to determne the equlbrum allocaton at pont a. Fgure 2 llustrates the consequences of the government ncreasng the health budget n the case where there s no change n the overall endowments. Inputs of sklled and unsklled * S H labour n the health sector ncrease to and U respectvely. The provson of extra health * H * S W care reduces the numbers on the watng lsts to and U. * W Fgure 2: Expanson of the Health sector wth unchanged endowments: example 1 The expanson of the health sector and the contracton of the watng lst changes both the total amounts of factors avalable to the two tradables sectors and the relatve sklledunsklled rato. It s convenent to decompose these nto a scale effect (ncreasng both effectve factor endowments by the lower of the two actual ncreases) and a factor-bas effect (ncreasng the effectve endowment of the hgher-growth factor). Splttng the changes nto the two components allows us to draw some nsghts from standard trade theory results. 13 Snce the health sector s, n ths example, the most skll-ntensve sector, ts expanson wll 13 These results have ther orgn n the semnal paper by Rybczynsk (1955). 8

9 lead to a contracton of the X sector due to the reducton n the avalablty of both factors, compounded by a reducton n the rato of sklled to unsklled workers avalable to the tradables sectors. The effect on the output of the Y sector wll be a mx of contracton due to the reduced avalablty of both labour types and expanson due to the reducton n the rato of sklled to unsklled workers. In the Fgure 2 example t s evdent that the net effect s a contracton of the X sector and an expanson of they sector. However t wll also be evdent that n general the effects on the tradables sectors depend on the orderng of factor ntenstes of the three producton sectors and the endowment rato, and on the ncdence of llness and of the provson of treatment for the two types of labour. For example, Fgure 3 shows a case where the health sector s smaller than the watng lst sector (both havng the same factor ntenstes as n Fgure 2), but wth the health sector havng much greater leverage on the sze of the watng lst sector. Fgure 3: Expanson of the Health sector wth unchanged endowments: example 2 Here the scale effect domnates the factor-bas effect: the former leads to an ncrease n the outputs of both tradable goods, and the ncrease n the output of the X good from the former s greater than the decrease n X output due to the factor-bas effect (whch also leads to a further ncrease n Y output). Thus outputs of both tradables ncrease. Whether the health sector s, n fact, more skll-ntensve than all other sectors s an emprcal queston, as s that of whether the ncdence of llness and the provson of health care are both 9

10 ndependent of labour type. In a mult-sectoral model wth more than two factors the foregong predctons are most unlkely to be wholly true. Nevertheless, they gve a useful gude to the nterpretaton of the outcomes of such a model. 4. The UK health system and health polcy The UK health system s domnated by state provson va the Natonal Health Servce (NHS) (devolved to regonal health authortes wth responsblty for hosptals, general practtoners and ancllary servces). There s some prvate provson va nsurance schemes that use prvate facltes, but also buy facltes and sklls from the state sector. Prvate provson s mostly of secondary (hosptal) care and covers approxmately 12 per cent of cases. (All NHS medcs workng n the secondary sector have contracts that allow them to provde prvate treatment, some 75 per cent dong so, and many NHS facltes are avalable for hre by prvate provders.) Fnancal provson for the NHS s set by the government over a fve-year plannng perod, and the responsble department (the Department of Health) must bd for a share of the overall budget n competton wth departments responsble for the armed forces, educaton, law enforcement etc.. The NHS admnstraton tself works to a rollng three-year plannng horzon, and may seek margnal adjustments to state fnance on an annual bass. UK governments (of both major poltcal partes) have been exercsed by the escalatng costs of the NHS. They have some control of some nputs, (e.g. salares of staff, workng practces, captal provson) but less control of others (e.g. pharmaceutcals prces). More mportantly, they are faced wth longer-term problems such as the ncreasng longevty of populaton and the demand for the use of new, and usually more costly, technologes. Evdence of pressure on secondary treatment facltes s provded prmarly by the length of the watng lsts for treatments the so-called ratonng by delay polcy operated by the NHS. Some non-lfe-threatenng complants are subject to long delays before treatment, and the exstence of such queues may n tself act as a dsncentve to seek treatment. As part of ts longer term strategy the NHS also ratons by denal n blockng, or at least delayng, the adopton of new technologes The fear s that f a new, superor, technology s approved then there wll be a sgnfcant ncrease n demand, whereas f t s not avalable then patents wll not mss what they don not have. Ths s analogous to the effects on recrutment of frng costs dscussed n the labour economcs lterature; see for example Garbald (1998). 10

11 1,400,000 1,200,000 1,000,000 Mllons watng 800, ,000 Total <3 months 3~5 months 6~11 months > 12 months 400, ,000 0 Jun- 98 Sep- 98 Dec- 98 Mar- 99 Jun- 99 Sep- 99 Dec- 99 Mar- 00 Jun- 00 Sep- 00 Dec- 00 Mar- 01 Jun- 01 Sep- 01 Dec- 01 Mar- 02 Jun- 02 Sep- 02 Dec- 02 Fgure 4: In-patent watng lsts (ordnary and day case admssons), England 15 Government polcy on medcal procedures and assocated nputs does not always focus on best provson, even wthn exstng technologes. For example, emprcal evdence suggests that some surgeons use procedures that le well above the lower envelope of exstng effcent cost-outcome procedures. 16 The UK government s unwllngness to tackle the more ntractable neffcences n the system (e.g. workng practces), coupled wth the (poltcal) necessty of restrctng costs, has led them to focus on more controllable costs. In parallel wth (generally successful) efforts to restrct medcal salares, mantan current workng practces, etc, the government has sought to lmt the ever-ncreasng cost of pharmaceutcals. 17 The chosen nstruments have been to negotate prce agreements wth provders and to requre doctors to prescrbe only from approved lsts of drugs 18 mostly generc products whch, beng out of patent, are cheaper (but arguably less effectve). Nevertheless, prces of pharmaceutcals are not wthn drect government control. Therefore, an ncrease n the prce of drugs poses problems for the government. The extreme optons are to: (a) mantan the offcal budget expendture, mplyng a reducton n expendture on other 15 Hosptal Inpatent Watng Lst Statstcs, England, NHS Trust based, the 'Green Book', avalable from (accessed ). 16 The Health Technology Assessment unt n the UK and the Natonal Insttute for Clncal Excellence n the USA seek to dentfy and promote best practce. 17 As do the governments of most countres. 18 The UK government operates a system of black, grey and whte lsts of pharmaceutcals. 11

12 treatment costs (numbers treated, tme spent n hosptal), or (b) to expand the budget at the cost of offsettng cuts elsewhere and so reductons n other government-provded goods. 5. The database and the CGE model The CGE model s a sngle-country comparatve statc model of the UK economy calbrated to 2000 data. The SAM underlyng the model s derved from the UK Supply and Use Balances for 2000, 19 other natonal accounts sources, and the General Household Survey (GHS) The structure of producton, output, demand and trade are taken from the supply and use balances, whch provde a commodty-by-ndustry use matrx for 123 commodtes and ndustres whch are aggregated to eleven sectors for the purposes of ths analyss (among whch publc and prvate health care, the pharmaceutcal ndustry and a sector producng medcal, precson and optcal nstruments). A commodty-by-ndustry make matrx s derved from data on ndustry and commodty output n 2000 and the most recent publshed make matrx for the UK, for The GHS s used to dsaggregate labour payments nto two types (sklled and unsklled) and to provde data on fve household types (pensoners, non-workng households wth and wthout chldren and workng households wth and wthout chldren). The UK Natonal Accounts Blue Book 21 s used to ensure that household aggregates are correct. The model employs standard assumptons regardng perfect competton, constant returns to scale and smultaneous market clearng. The Harberger conventon s used throughout, so that the model s calbrated so that all prces are equal to one n the benchmark. Producton and consumpton behavour are modelled usng the constant elastcty of substtuton (CES) famly of functons, whch ncludes Leontef, Cobb-Douglas and constant elastcty of transformaton (CET) functons. Cross-border trade s treated under a set of commonly used assumptons. Frstly, the UK s treated as a small open economy, mplyng that the UK faces exogenous world prces for mports and exports. Secondly, products are dfferentated 19 Avalable from the homepage of the Offce for Natonal Statstcs (ONS), Natonal Statstcs Onlne: (accessed ). 20 The GHS s carred out by the Socal Survey Dvson of the ONS collectng nformaton on a range of topcs from people lvng n prvate households n Great Brtan. The complete dataset s avalable from the UK Data Archve (UKDA): (accessed ). The ONS and UKDA bear no responsblty for the analyss or nterpretaton of the data as lad out n ths paper. 21 Avalable from Natonal Statstcs Onlne: (accessed ). 12

13 accordng to regon of orgn accordng to the famlar Armngton assumpton, so that mports and exports are qualtatvely dfferent from domestcally produced goods. It seems a reasonable smplfcaton to model health provson as a non-tradable output (usng traded ntermedates) that adds value to the ll, who are treated as an ntermedate nput. Fgure 5: Modellng Health Provson The gven health budget lmts nputs of factor servces and ntermedates, usually to a level nsuffcent to treat all those presentng themselves as ll. The output WELL s thus, n terms of people, less than the nput ILL. The WELL output could be vewed n two ways: (a) t s the number of people treated and cured, the remander beng added to the watng lst; or (b) t s the proportonal reducton n the degree of llness of all groups, wth the proporton of semcured workers becomng an addton to the effectve workforce. In ths paper we report on a comparatve statcs analyss, the easest formulaton. The obvous gans from ths approach are that we need not model longer-term populaton processes (brths, deaths, transtons from YOUNG to WORKING and from WORKING to RETIRED), nor do we have to model the decomposton of those movng from YOUNG to WORKING nto SKILLED and UNSKILLED. The major dsadvantage s that we have to translate the health transton from ILL to WELL, whch s less than 100% n a dynamc framework, nto a one-perod model. However, an offsettng advantage s that we can gan nsghts nto the mplcatons of polcy changes from our earler low-dmenson analytcal model. 13

14 We model the nteracton between health and labour supply n the statc CGE model by the use of a non-partcpaton rate η for each type of labour. Non-partcpaton can be nterpreted as beng on the watng lst, whereas partcpaton mples employment n one of the sectors of the economy. The effectve labour supply and the non-partcpaton rate: ( 1 η ) LS LE LS s then a functon of labour endowments = 0 η 1 F (3) The non-partcpaton rate η, s formulated as a constant elastcty functon of the level of health, H: η β = γ H β 0, γ 0 LE > F (4) γ s calbrated so that the benchmark levels of health provson lead to the requred nonpartcpaton rate. 22 It measures the effectveness of the level of health, such that a reducton n γ, gven H, reduces the non-partcpaton rate and ncreases effectve labour supply. The parameter β s an elastcty that determnes how non-partcpaton rates respond to changes n health provson (.e. t measures the effectveness of changes n health). β = 0 mples that no such response occurs, and wll be used as one of a number of possble values of β n the reported smulatons. The level of health depends on the level of publc and prvate health provson Q and NHS Q PHC : Fgure 6: The nested producton functon Output Qj, j G at prce PP j Output at producer prces Output tax t j Cobb-Douglas a Value Added Demand for Intermedate Goods X, j P, G Captal K j R j Labour L j W j 14

15 H α ( 1 α ) = QNHS QPHC 0 α 1 F (5) α s a parameter specfyng, for each labour type, what proporton of health s determned by publc as opposed to prvate health provson. 23 The structure of producton n each of the eleven sectors s shown n Fgure 6. For each sector n the set of produced goods G, a mult-stage producton functon s specfed n equatons (6). The zero-proft condton s gven n equaton (7). Producton s taxed at an ad valorem rate t j, and s a Leontef functon of value added and nputs of ntermedate goods. Value added s a Cobb-Douglas functon of labour and captal nputs. 1 Qj = mn CD( E, j F), X, j for G j G (6) ( 1 t j ) 1 d QPP j j = E, jw + X, P t F G ( 1 j ) j j G (7) Output of each sector Q s transformed nto supply of each commodty as shown n Fgure 7. j Equaton (8) shows that producton by sector j s a CET functon of the producton of ndvdual commodtes by sector j, commodty supply. Q (, ; ) Q, j. Equaton (9) gves the zero-proft condton for = CET Q τ j G (8) j j j Fgure 7: Output transformaton Domestc Supply of commodty ( G) by ndustry j Q, j P, j G Output of ndustry j Q j PP j, j G CET (elastcty τ j ) 22 The non-partcpaton rate for sklled and unsklled labour n the benchmark s 10%. 15

16 QPP = Q P j G (9) j j, j G m m Fgure 8 shows the structure of market supply. Imports Q and domestc supply Q (whch s the sum of each ndustry j s output of commodty, Q ) are aggregated accordng to the Armngton assumpton (goods are dfferentated accordng to ther orgn), wth a CES functon defnng the substtuton possbltes between mports and domestc goods. Aggregate supply s then dvded nto export supply d Q, j x Q and supply to the domestc market (equatons (10) and 9). Equatons 10 and 11 show how domestc market supply s transformed nto aggregate supply at basc prces by the addton of transport and trade margns. Equaton (14) s the equlbrum condton n goods market, where aggregate supply s equal to the sum of ntermedate demands from dfferent ndustres plus the sum of demands by dfferent households, demand by the government, export demand and nvestment demand. Fgure 8: Market Supply a a Aggregate supply Q P, G Aggregate supply at basc prces Consumpton tax c Producton for export market Q x x P Domestc market supply Q d P d Margns X P m margn Armngton aggregate supply CET (elastcty τ j ) CES ( φ ) m Imports Q P m Domestc supply Q P Output of commodty by ndustry j Q P j G, j d, 23 We assume that unsklled labour (.e. those on low ncome) has access to publc health care only, α unsklled = 1. 16

17 x d m CET ( Q, Q ; τ ) = CES ( Q, Q ; φ) (10) x x d d m m + = + Q P Q P Q P Q P (11) Q a 1 mn d, m = Q 1 c X G (12) 1 QP = QP + X P a a d d m marg n ( 1 c ), Q = X + X + X + X + X a, j, j, GOV, XPT INV j G j H G G (13) (14) Sx representatve agents n the model consume goods and servces: fve prvate households and government. The prvate households (denoted by the set H) and government are demonstrated n Fgure 9. Each of the fve households receves ncome from captal (beng the sum of ncomes from a varety of sources, from ownershp of frms to nterest on savngs or pensons), and government transfers. The pensoner household and the two workng households receve ncome from employment, n proporton to ther earnngs from unsklled and sklled labour endowments. Each household pays drect taxes on labour and captal ncome, and spends ts dsposable ncome on the eleven commodtes and on savngs. Equaton (15) gves prvate dsposable ncome Y. ( ) Y = LSj, Wj 1 tj +T H j F (15) The remander of publc and all prvate health care s consumed by sklled labour (α sklled = 0.84). 17

18 Equaton (16) gves the defnton of prvate expendture E as a functon of prvate savngs and prvate expendtures. The real level of savngs s gven n equaton (18) as a functon of demands for savngs, whch s equal to output of nvestment, a functon of nvestment demands for ndvdual commodtes. Equaton (19) equates the value of savngs wth the value of nvestment demand for ndvdual commodtes. Equaton (20) equates net ncome wth expendture. E = Q P + X Pj H s s j, j G (16) ( XSAV, X j, ; ) U = CES j G ρ H (17) SAV, (, INV G) s Q = X = f X for H s s QP XINV, P G = (18) (19) Y = E H (20) The government receves ncome from ndrect and drect taxaton, consumes three commodtes ( health, publc admnstraton and defence and other servces ) and purchases a fxed small amount of foregn exchange n order to fnance the trade surplus. It Fgure 9: The prvate household and government Factor ncome Drect taxaton Indrect taxaton Prvate Households Transfers Government Trade balance Consumpton Prvate consumpton Investment Government consumpton 18

19 transfers the remander of ts ncome to the fve households, n proporton to ther orgnal ncome from government benefts n the orgnal data. Ths consttutes the government closure rule n the model. Other closure rules are that net foregn nvestment s fxed, and therefore the trade balance s also fxed, and nvestment s savngs-drven. The model s solved n MPSGE wth the exchange rate set as the numérare. Welfare results are derved for each household as the equvalent varaton of prvate household utlty U. Overall welfare changes are computed as the sum of household equvalent varatons plus the change n the government s provson of publc goods (health servces). 6. The counterfactual smulatons In order to llustrate the functonng of the model and potental applcaton areas we perform two smulatons. Frstly, we consder the effects of a 10% ncrease n expendtures on the NHS n a scenaro of perfectly moble factors of producton (experment 1). The expanson of publc health care, although drawng away resources from other sectors n the economy, s expected to mprove health, labour market partcpaton and welfare. Secondly, we smulate a 10% ncrease n the market prce of pharmaceutcals under a scenaro n whch the government keeps ts (overall) health budget fxed n value (experment 2) and one where the government ncreases expendtures on NHS health care, such that total real government expendtures on health care are unchanged (experment 3). In the frst scenaro, the ncrease n the cost of pharmaceutcal nputs mples lower levels of health care provsonng and that less people are recevng treatment, and, hence, deteroraton n health, lower labour market partcpaton rates, lower (workng) households ncomes and a fall n overall welfare. We expect that negatve effects on health, labour market partcpaton and welfare wll be mtgated n the second scenaro as the government ncreases expendtures on health care and so mantans the prevous level of health care provsonng,.e. number of treatments, under the NHS. A senstvty analyss s carred out to check the robustness of results for the health elastcty parameter β. Results are reported for β=0 unless ndcated otherwse; these experments gve the drect effects of addtonal expendture on health sector employment and output but suppress the ndrect effects comng through, for example, ncreased labour market partcpaton. 19

20 Experment 1: a 10% ncrease n publc health expendtures perfectly moble factors Government expendtures are fxed n foregn exchange terms, so that the mmedate effect of an expanson of publc health expendtures n the face of gven tax revenues s to reduce transfer payments to households by 5.17%. The addtonal NHS resources result n an ncrease n publc health care provson by 9.94% and, va nput-output lnkages, ncrease the demands and producton of pharmaceutcal products (by 5.01%) and medcal, precson and optcal nstruments (by 1.71%). As a result health care, pharmaceutcals and nstruments become relatvely more expensve (domestc prces ncrease by 0.055%, 0.035% and 0.043% respectvely,) whch ncreases the costs and hence reduces the sze of prvate health care provson (by 0.94%), all remanng sectors contract. In the trade balance, the major changes occur n the health-related sectors: mports of pharmaceutcals rse by 4.69%, exports rse by 4.61%. Imports and exports of medcal, precson and optcal nstruments ncrease by 1.61% and 1.51% respectvely. The ncrease n publc health care boosts the health of unsklled labour (and hence ts partcpaton rate) by 9.94%. Ths s relatvely more than the mprovement of 8.08% n the health of sklled labour. Unsklled labour s health s affected only by changes n publc health care whereas sklled labour s health s affected by changes n both publc and prvate health care provson (Table 1). Table 1 Changes n health, experment 1 (%) Type of labour β =0 β=0.5 β=1 β=2 Sklled labour Unsklled labour As the health elastcty, β, s ncreased, the mproved health of workers feeds back nto the labour market va mproved labour market partcpaton rates (Table 2). Agan, unsklled labour s able to beneft slghtly more. 20

21 Table 2 Changes n labour market partcpaton rates, experment 1 (%) Type of labour β =0 β=0.5 β=1 β=2 Sklled labour Unsklled labour For β = 0, sklled labour gans and unsklled labour and captal lose from an expanson n (publc) health care and contracton of non-health sectors. As β rses both sklled and unsklled labour are worse off (Table 3). Table 3 Changes n factor rewards, experment 1 (%) Type of factor β =0 β =0.5 β=1 β=2 Sklled labour Unsklled labour Captal The changes n factor remuneraton can be explaned n terms of the changes n labour market partcpaton rates. As the health elastcty goes up, an gven ncrease n the level health leads to a hgher labour market partcpaton rate. Labour thus becomes relatvely less scarce compared to captal, whch s reflected n a fall n wages and a rse n rents. At low levels of β, all households are relatvely worse off, except for chldless workng households who beneft from the rse n sklled wages (Table 4 and 5). As β rses, workng households utlty eventually mproves as these households beneft from ncreased labour market partcpaton. As β s ncreased, producton n all sectors eventually rses. Ths boosts government tax revenues up to 0.95% for β=2 and mtgates the fall n transfers to all households (transfers fall by 2.27% for β =2). Consequently, non-workng households are relatvely less worse off as β rses. The mproved welfare of workng households (who beneft drectly from ncreased partcpaton n the labour market), combned wth smaller reductons n the welfare of non-workng households (for whom the fall n government transfers s less pronounced) and ncreases n government consumpton of health and non-health goods, can lead to an overall welfare gan of up to bllon pounds (a relatve change of 1.07%). 21

22 Table 4 Equvalent varatons, experment 1 (bllons ) Type of household β=0 β=0.5 β=1 β=2 Pensoners Non-workng, chldren Non-workng, no chldren Workng, chldren Workng, no chldren Overall (ncl. government consumpton of publc goods) Table 5 Equvalent varatons, experment 1 (% of orgnal ncome) Type of household β=0 β =0.5 β=1 β=2 Pensoners Non-workng, chldren Non-workng, no chldren Workng, chldren Workng, no chldren Overall (ncl. government consumpton of publc goods) Experment 2: A 10% rse n the prce of pharmaceutcals government budget fxed n value Government expendtures on health care, ncludng all pharmaceutcal mports, are fxed n value (foregn exchange) terms such that a rse n the prce of health care lowers the quantty of health care provson. Smlarly, a rse n the prce of pharmaceutcal mports s matched by and ncrease n foregn exchange needed to pay for these mports, such that pharmaceutcal mports are constant. The model returns values for the domestc prce (rses by 18.6%) and the 22

23 mport prce (ncreases by 2.07%), whch agree wth the targeted ncrease n the prce of the pharmaceutcal composte of 10%. 24 Followng the pharmaceutcal prce rse resources move nto the producton of pharmaceutcals (grows by 1.4%) and consumpton falls (by approxmately 9.4%). Imports of pharmaceutcals are constant by desgn such that exports contract by approxmately 19.5%. Confronted wth hgher unt costs of ntermedate nputs, publc and prvate health care ncrease ther prces by approxmately 0.86%. The rse n the prce of health care leads to a drop n demand for prvate health care and a fall n the producton of publc health care (NHS and prvate health care contract by 0.85% and 1.01% respectvely). Producton, consumpton mports and exports and prces of other sectors reman farly constant. All factors of producton experence a fall n relatve rewards of less than 0.26% for all values of the health elastcty. As predcted, the government consumes and provdes less health care (a fall of approxmately 0.85%) and ncreases producton of other goods (other servces and publc admnstraton and defence ncrease by less than 0.2%). Lower levels of publc and prvate health care mply a fall n the level of health for sklled and unsklled labour. On average health deterorates by 0.86%, but the change s slghtly more pronounced for sklled labour as they consume both publc and prvate health care, and the latter contracts relatvely more. Consequently, labour market partcpaton rates fall a for postve health elastcty (agan by more for sklled labour). Government ncome from taxes falls by 0.11%. In order to keep ts fnances balanced the government reduces transfers to households by 0.28%. Household welfare dmnshes by a farly small percentage for workng households (<0.25% across all levels of β), but by relatvely more for non-workng households who are solely relant on ncome from government transfers. Ths mples that the loss n wage ncome s relatvely less than the loss n ncome from transfers such that households who receve ncome from both sources are less worse off than households lvng from state benefts only. Overall welfare (a composte of household expendture on consumpton and savngs, and government consumpton) deterorates by around bllon pounds. 24 We obtan a 10% ncrease n the market prce of pharmaceutcals by postulatng that the prce change s drven by an ncrease n the exogenous world mport prce of pharmaceutcals, whch then generates the mentoned changes n mport and domestc pharmaceutcal prces. 23

24 The results are relatvely robust to changes n the elastcty of labour market partcpaton wth respect to health (β). The effect of changng the health elastcty s llustrated below for a selecton of varables. Table 6 Change n labour market partcpaton rates: experment 2 (%) Type of labour β=0 β =0.5 β =1 β =2 Sklled labour Unsklled labour Table 7 Changes n health: experment 2 (%) Type of labour β=0 β =0.5 β=1 β =2 Sklled labour Unsklled labour Table 8 Changes n factor rewards: experment 2 (%) Type of factor β=0 β =0.5 β=1 β =2 Sklled labour Unsklled labour Captal For a gven fall n the level of health, a hgher β exacerbates the fall n labour market partcpaton rates (Table 6). A lower labour market partcpaton rate decreases effectve labour supply to all sectors n the economy, ncludng health care, thereby renforcng the negatve health effect of lower levels of health care. As a consequence, the fall n health of workers s more pronounced (Table 7). Also, gven the demand for factors of producton, the more pronounced fall n effectve labour supply for hgher values of the health elastcty mples that relatve to captal, labour becomes relatvely scarce. Ths renforces the fall n rents on captal and mtgates the fall n wages (Table 8). Fnally, a hgher elastcty of labour market partcpaton to health ncreases the loss n household welfare as measured by the 24

25 equvalent varaton (Table 9). Relatve to orgnal ncome non-workng households are stll worse off than workng households due to the fall n ncome transfers as β rses (Table 10). Table 9 Equvalent varatons: experment 2 (bllons ) Type of household β=0 β =0.5 β=1 β=2 Pensoners Non-workng, chldren Non-workng, no chldren Workng, chldren Workng, no chldren Overall (ncl. government consumpton of publc goods) Table 10 Equvalent varatons: experment 2 (% of orgnal ncome) Type of household β=0 β =0.5 β=1 β=2 Pensoners Non-workng, chldren Non-workng, no chldren Workng, chldren Workng, no chldren Overall (ncl. government consumpton of publc goods) Experment 3: A 10% rse n the prce of pharmaceutcals level of publc health care fxed In addton to experment 2, the government ncreases ts expendture on NHS care n order to mantan the number of people that are recevng treatment under the NHS. From the prevous experment we know that the prce of publc health care ncreases by approxmately 0.86% followng a 10% rse n the prce of pharmaceutcals. Hence, n order to mantan prevous 25

26 levels of publc health care the government should match the prce ncrease of health care by an dentcal ncrease n expendtures on publc health care of 0.86%. As ths s such a mnor change, results dffer only margnally and a short summary of the dfferences s gven below. Pharmaceutcal producton goes up by 0.47% extra as more ntermedate nputs from the pharmaceutcal ndustry are needed to produce the addtonal publc health care. For dentcal reasons, producton of medcal, precson and optcal nstruments expands by 0.19% (compared to 0.04% n experment 2). By assumpton, the level of publc health care remans constant, such that prvate health care contracts by 1.09%, 0.08% more than n experment 2. A dstnctve result s that the health of unsklled labour, whch depends solely on levels of NHS provsonng, s mantaned at ts orgnal level. Therefore, unsklled labour partcpates at the same rate as before and supples the same amount of labour. Sklled labour s worse off n terms of health by 0.18% (though better off than n experment 2) and n terms of labour market partcpaton (falls n range of 0.01% to 0.04% for postve health elastcty). In order to mantan levels of publc health care followng an ncrease n the cost of provsonng, the government must reduce transfer payments to households, gven that expendtures on other goods are stll fxed n value. Consequently, compared to experment 2 all households are worse off for low levels of β. As β rses, workng households are better off relatve to experment 2 because of mproved health and hgher labour market partcpaton. In lne wth these patterns, overall welfare falls by and bllon pounds for β = 0 and β = 2 respectvely (the former decrease s hgher and the latter decrease s lower compared to experment 2). 7. Conclusons and comments Ths paper presents the results from a prelmnary CGE assessment of health care n the Unted Kngdom. The paper models health provson and ts effects on effectve labour supply, and shows the magntude of changes to factor rewards, health and labour market partcpaton, and welfare that could be expected from scenaros of ncreased publc health expendture and ncreases n pharmaceutcal prces. The man fndngs of the paper are that ncreased publc health expendture mproves health and leads to hgher labour partcpaton for both sklled and unsklled labour. These results depend on the effectveness of health care, as represented n the model by an elastcty governng how labour partcpaton rates respond to health provson. A farly low rate of effectveness s enough to ensure a welfare gan from ncreased publc health provson. 26