ArticlePDF Available

The Relationship between Health Care Expenditure and Health Outcomes

Authors:

Abstract and Figures

The relationship between health care expenditure and health outcomes is of interest to policy makers in the light of steady increases in health care spending for most industrialised countries. However, establishing causal relationships is complex because, firstly, health care expenditure is only one of many quantitative and qualitative factors that contribute to health outcomes, and, secondly, measurement of health status is an imperfect process. This study reviews key findings and methodological approaches in this field and reports the results of our own empirical study of countries of the European Union. Our analysis examines life expectancy and infant mortality as the 'output' of the health care system, and various life-style, environmental and occupational factors as 'inputs'. Econometric analyses using a fixed effects model are conducted on a panel data set for the former 15 members of the European Union over the period 1980-1995. The findings show that increases in health care expenditure are significantly associated with large improvements in infant mortality but only marginally in relation to life expectancy. The findings are generally consistent with those of several previous studies. Caveats and improvements for future research are presented.
Content may be subject to copyright.
Orig i nal Pa pers
John Nixon
1
· Phil ippe Ul mann
2
1
Cen tre for Re views and Dis sem i na tion, Uni ver si ty of York, UK
2
Chair of Eco nomics and Man age ment of Health Ser vices,
Con ser va toire Na tion al des Ar ts et Métiers, Par is, France
The re la tion ship be tween
health care ex pen di ture
and health out comes
Ev i dence and caveats for a caus al link
Eur J Health Econom 2006 · 7:7–18
DOI 10.1007/s10198-005-0336-8
Pub lished on line: 21. Jan uary 2006
© Springer Medi zin Ver lag 2006
Evi dence con firms steady trends in in-
dus tri alised coun tries for im proved health
out comes and for in creas ing health ex pen-
di ture. In the coun tries of the Eu ro pean
Union (EU) be tween 1960 and 1995, for
ex am ple, the av er age in fant mor tal i ty ra-
te fell from 3.3 to 0.6 deaths per 1,000 li-
ve births; av er age life ex pect an cy at birth
for fe males rose from 72.5 to 80 years; av er-
age male life ex pect an cy at birth rose from
67.6 to 73.6 years; whilst over the same pe-
ri od to tal health ex pen di ture as a share of
gross do mes tic prod uct (GDP) rose from
3.4 to 7.7 [29].
How ev er, the ev i dence for a caus al
link be tween health care ex pen di ture and
health out comes re mains elu sive as prob-
lems emerge from ‘the dif fi cul ty of iso lat-
ing the con tri bu tion of the health ser vice
in put as a de ter mi nant of health sta tus
out put”... which frus trates at tempts to mea-
sure the over all ef fec tive ness and ef fi cien-
cy of health care’ [13]. Con se quent ly in
com par ing trends with in a coun try over
time (e.g. for lon gev i ty) there is no ex per i-
men tal con trol group pro vid ing com pa ra-
ble data in the ab sence of health ser vices.
More over, in most cas es an in di vid u als
vis it to a health pro fes sion al for treat ment
(e.g. a den tist or a chi rop o dist) in volves
health care ex pen di ture but does not nec-
es sar i ly re sult in ex tend ing his or her life
span; it sim ply brings some im prove ment
in the in di vid u als feel ing of ‘well be ing.
This is one of the rea sons for the in tro duc-
tion of util i ty mea sure ments in health ca-
re to de rive qual i ta tive out come mea sures
such as the qual i ty-ad just ed life year (QA-
LY) and heal thy years equiv a lent (HYE)
[26] and the Health Util i ties In dex (HUI)
[20], which, how ev er, in volve a num ber
of un re solved is sues as so ci at ed with def i-
ni tions, mea sure ments and prac ti cal use.
There fore rel a tive ly few stud ies have been
suc cess ful in find ing a link be tween health
care ex pen di ture and health out comes, as
oth er fac tors af fect ing health out comes
such as diet, life-style and en vi ron ment
are of ten tak en to be the prin ci pal fac tors
af fect ing health out comes, and par tic u lar-
ly life ex pect an cy.
The aim of this study is to ex am ine this
re la tion ship fur ther. The con tri bu tion be-
gins with an overview of com mon ly adopt-
ed ap proach es by re searchers in this field
of study, fol lowed by a re view and sum ma-
ry of the find ings of key stud ies in this area
of re search. We then de scribe the meth-
ods that we have adopt ed to un der take
our own em pir i cal anal y sis of the re la tion-
ship be tween to tal health care ex pen di ture
and health out comes in the coun tries of
the EU over the pe ri od 1980–1995, us ing
life ex pect an cy (fe males and males) and in-
fant mor tal i ty as the de pen dent vari ables.
The re sults are then pre sent ed, fol lowed
by a dis cus sion of the caveats and lim i ta-
tions of the work and pre vi ous stud ies in-
clud ed in the re view, and how fu ture stud-
ies might be im proved.
Method olog i cal ap proach es
In this field of en quiry it is pos si ble to dis-
tin guish two dis tinc tive ap proach es that ha-
ve been adopt ed by oth er re searchers [44].
The first ap proach is ground ed in the work
of Gross mans [14] hu man cap i tal the o ry at
the lev el of the in di vid u al, which re gards
health as a com mod i ty which the in di vid u-
al will wish to con sume and max imise, sub-
ject to his bud get con straints, in con junc-
tion with a num ber of en dog e nous and ex og-
e nous vari ables or char ac ter is tics which ha-
ve an im pact on an in di vid u als health. With-
in this mod el, in come and ed u ca tion al lev-
el play promi nent roles as ex plana to ry vari-
ables. Gross mans house hold pro duc tion
func tion mod el of con sumer be hav iour was
fur ther de vel oped to ac count for the gap be-
tween health and med i cal care as one of the
many in puts into its pro duc tion. The ‘in vest-
ment mod el of de mand’ deals with a the o-
ret i cal and em pir i cal in ves ti ga tion of the de-
mand for the com mod i ty ‘good health’ [15].
The mod el es sen tial ly re gards health as a cap-
i tal good that is in her it ed and de pre ci ates
or de te ri o rates over time. The the o ry po-
sits that in vest ment in health is a pro cess in
which med i cal care is com bined with oth-
er rel e vant fac tors to pro duce new health,
which, in part, off sets the pro cess of de te ri o-
ra tion in health stock. The pos i tive cor re la-
tion be tween ed u ca tion and the health of an
in di vid u al has been con firmed in many sub-
se quent stud ies [15].
7Eur J Health Econom 1 · 2006
|
in come (and health ex pen di ture as one of
the study’s de pen dent vari ables). Health
ex pen di ture as a share of GDP, gross na-
tion al prod uct (GNP) or per cap i ta health
ex pen di ture are com mon ly used. In ad-
dress ing prob lems of po or com pa ra bil i ty
when us ing ex change rates, sev er al stud ies
adopt ed pur chas ing pow er par i ty (PPP) [3,
5, 27, 30] when us ing health ex pen di ture
or in come data. Ten of the 15 stud ies use
in come as an ex plana to ry vari able in ad-
di tion to health ex pen di ture, but there is
a need to ac knowl edge, as sev er al stud ies
do, that the cor re la tion be tween these two
vari ables is high [28].
In terms of oth er diet, so cio-eco nom-
ic and life-style ex plana to ry vari ables it is
pos si ble to ob serve wide vari a tions in the
num ber used (mean 8, range 3–17). The
stud ies util is ing the most ex ten sive num-
ber of ex plana to ry vari ables are Cochrane
et al. [8], which used sev en ‘health care
vari ables (such as physi cians, nurs es, beds
etc), six di e tary con sump tion vari ables
(in clud ing those com mon ly used in sev-
er al stud ies, e.g. al co hol con sump tion, to-
bac co, fat in take) and four de mo graph ic
and eco nom ic vari ables. Berg er and Mess-
er [7] used 12 ex plana to ry vari ables. Some
of the cho sen ex plana to ry vari ables also
clear ly re flect the hy poth e sis be ing in ves ti-
gat ed, as ex em pli fied by the use of pri vate
and pub lic splits in health care ex pen di-
tures [9] and ex pen di ture on phar ma ceu-
ti cals [3, 22, 27, 34]. To cap ture pop u la tion
ef fects a num ber of stud ies in clude an age-
spe cif ic vari able [3, 17, 21, 43] or a pop u la-
tion den si ty vari able [8, 9, 10]. Some ve-
ry spe cif ic ex plana to ry vari ables in clud-
ed in the sam ple are de cen tral i sa tion co-
ef fi cient, po lit i cal rights [32], pro por tion
of white-col lar work ers [30] and Na tion al
Health Ser vice (NHS) fi nanc ing of med i-
cal ser vices [21].
When ex am in ing the coun tries that
were stud ied, it can be seen from
. Ta-
ble 1 that the vast ma jor i ty (ten) stud ied
var i ous com bi na tions of OECD coun tries.
An oth er three in clud ed de vel oped coun-
tries or coun tries in West ern Eu rope, one
study in clud ed both de vel op ing and de vel-
oped coun tries, and two stud ies by the sa-
me au thor anal y sed the Cana di an provin-
ces [9, 10]. It is in ter est ing to note here that
these two stud ies used data with a high
de gree of ho mo ge ne ity and con sis ten cy,
out comes. We also, how ev er, ex am ined pa-
pers that ad dressed health in re la tion to
eco nom ic growth. (b) Pa pers pre sent ing
em pir i cal re sults ob tained from mac ro-
econo met ric mod els. (c) Pa pers ex plor-
ing these is sues at least for Eu ro pean coun-
tries or less specif i cal ly OECD coun tries.
(d) Stud ies based on larg er sam ples, i.e.
those in clud ing de vel op ing or tran si tion al
economies, were also con sid ered if they in-
clud ed Eu ro pean/OECD coun tries in the
sam ple. By adopt ing this pro ce dure we ini-
tial ly iden ti fied 38 po ten tial ly rel e vant pa-
pers. On clos er ex am i na tion 22 were delet-
ed ei ther be cause they did not meet our
in clu sion cri te ria, or be cause dif fer ent ver-
sions of same pa per (with mar gin al dif fer-
ences) ex ist ed. We fi nal ly re tained 16 pa-
pers for re view/sum ma ry. This ap proach
was found to be use ful as it fa cil i tat ed an
overview of the meth ods that have been
used, along with a sum ma ry of the prin ci-
pal re sults. For ease of as sim i la tion the rel-
e vant data for each study are sum marised
in
. Ta ble 1, which pro vides de tails of
the de pen dent vari able(s), ex plana to ry
vari able(s), the coun tries stud ied, a brief
de scrip tion of the mod el, and the prin ci-
pal re sults of each in clud ed study. Whilst
we ac knowl edge the wealth of lit er a ture
that has ex am ined vari a tions in health out-
comes re lat ed to in come in equal i ty (e.g.
[38, 39, 41]), we have not in clud ed this ty-
pe of anal y sis in the re view as it falls out-
side the scope of this pa per.
In terms of de pen dent vari ables the vast
ma jor i ty of stud ies uti lise mor tal i ty rates
(age-spe cif ic or in fant mor tal i ty in par tic-
u lar) and/or life ex pect an cy. Life ex pect an-
cy is used main ly at birth, but sev er al stud-
ies as sess life ex pect an cy ac cord ing to gen-
der and at spe cif ic ages oth er than birth
(e.g. 40, 69, 80 years). One study [27], how-
ev er, did use a health util i ty vari able (dis-
abil i ty-ad just ed life ex pect an cy or DALE)
at birth and at age 60 years, as well as po-
ten tial years of life lost for cir cu la to ry dis-
ease, can cer and res pi ra to ry dis ease.
All stud ies in clud ed some form of
health ex pen di ture as one of the de pen-
dent vari ables for the mod el used, with
the ex cep tion of Robali no et al. [32], which
was re tained for in ter est and com ment due
to its as sess ment of the im pact of fis cal de-
cen tral i sa tion on in fant mor tal i ty, and
Grubaugh and San terre [16], which used
The sec ond ap proach, adopt ed in this
study, con sid ers health as a pro duc tion
func tion which is ad dressed us ing ag gre-
gate or mac ro-lev el data. The ba sic tenets
of this ap proach are that health can be
viewed as an ‘out put’, say of a health care
sys tem, which is in flu enced by the ‘in puts
to that sys tem. In par tic u lar re searchers
adopt ing this ap proach wish to in ves ti-
gate the re la tion ship be tween health care
ex pen di ture, or med i cal care re sources as
in puts, and health out comes as the out put
of that sys tem. Fur ther more, this is sue has
be come a cen tral ques tion in the con text
of health care cost-con tain ment in most
de vel oped coun tries in the past few de-
cades.
How ev er, it is the case that the dis tinc-
tion be tween these ap proach es has be-
come some what blurred, and there is a de-
gree of over lap as many of the vari ables em-
ployed in the two ap proach es are the same,
and they are both cat e gorised as ‘pro duc-
tion func tions. For the pur pos es of our em-
pir i cal anal y sis we fo cus our at ten tion on
the sec ond ap proach due to the adop tion
of mac ro-lev el vari ables in our pro duc tion
func tion. More over, Ar row’s im pos si bil i ty
the o rem high lights the metho log i cal prob-
lems in at tempt ing to move from the mi-
cro-lev el to the mac ro lev el [2]. This is par-
tic u lar ly true in the health sec tor where
many re sults con firm health as be ing a lux-
u ry good at a mac ro lev el, when it is a nor-
mal good on a mi cro-lev el. Fur ther more,
anal y sis of this is sue has dem on stra ted an-
a lyt i cal ly and em pir i cal ly that con sid er ing
mi cro-lev el re sults for health pol i cy de ci-
sion mak ing at the mac ro-lev el may be
mis lead ing [31].
Pre vi ous re search
This sec tion pro vides a re view of key stud-
ies that have con sid ered the re la tion ship
be tween health ex pen di ture, among oth-
er ex plana to ry vari ables, and health out-
comes, us ing mac ro-lev el data. In or der
to iden ti fy po ten tial ly suit able stud ies we
first searched for all the po ten tial rel e vant
pa pers on these top ics us ing the In ter net
(pub lished ar ti cles, work ing pa pers, pub lic
re ports) or ar ti cles bib li o graph ic de tails.
Our in clu sion cri te ria were: (a) Pa pers
had to fo cus main ly on the re la tion ship be-
tween health care ex pen di tures and health
8
|
Eur J Health Econom 1 · 2006
Original Papers
which the au thors in di cate is an im por tant
fac tor in ob tain ing their re sults (in favour
of a link be tween health ex pen di ture and
health out comes). In con sis ten cies in oth-
er datasets, they ar gue, of ten con tribute
to neg a tive find ings when ex plor ing such
links. It is re cog ni sed that the in clu sion
of the study that ex am ined both de vel op-
ing and de vel oped coun tries [5] may cre-
ate prob lems con cern ing the com pa ra bil i-
ty of stud ies, but this study was re tained as
it of fers some in sights into oth er ex plana-
to ry vari ables that are rel e vant to de vel op-
ing coun tries and fac tors as so ci at ed with
ge o graph i cal lo ca tion.
When con sid er ing the var i ous mod-
elling tech niques that were adopt ed, it is
pos si ble to dis tin guish a num ber of fea-
tures. As would be ex pect ed, all stud ies
utilised some form of mul ti var i ate re gres-
sion anal y sis, with some in cor po rat ing
lagged vari ables for data af fect ed by tem-
po ral fac tors [21, 22, 27, 34]. The sec ond re-
lates to the form of data anal y sis used; ten
used pan el data sets (time-se ries cross-sec-
tion al), three used cross-sec tion al data for
more than 1 year, and three used cross-sec-
tion al data for only a sin gle year. The third
point of in ter est is the trans for ma tion of
data into logs by sev er al stud ies to en able
the elas tic i ties of sig nif i cant ex plana to ry
vari ables to be used. In some cas es the
mod elling in cor po rat ed shift dum mies to
ac count for fixed ef fects with in the sam ple,
for ex am ple, in in ves ti gat ing het ero ge ne ity
due to coun try-spe cif ic ef fects [17] or the
im pact of health care sys tem (NHS or so-
cial in sur ance) [11].
The prin ci pal re sults showed that
health ex pen di ture was a sig nif i cant ex-
plana to ry vari able for at least one health
out come ex am ined in 12 of 16 stud ies.
Five stud ies found that in come was a sig-
nif i cant ex plana to ry vari able. One stu-
dy found that fis cal de cen tral i sa tion leads
to a de crease in in fant mor tal i ty rate [32],
and one study did not find health ex pen di-
ture to be sig nif i cant when con trol ling for
in come [21]. It is in ter est ing to note that
all stud ies that in clud ed phar ma ceu ti cal
ex pen di ture found this as pect of health ex-
pen di ture to be sig nif i cant and pos i tive for
health out comes [9, 22, 27, 34]. In terms
of life-style vari ables a num ber of stud ies
found that smok ing [7, 8, 9, 10, 16, 43], al co-
hol [7, 8, 9, 10, 16, 30] and con sump tion of
Abstract
Eur J Health Econom 2006 · 7:7–18
DOI 10.1007/s10198-005-0336-8
© Springer Medi zin Ver lag 2006
John Nixon · Phil ippe Ul mann
The re la tion ship be tween
health care ex pen di ture
and health out comes. Ev i dence
and caveats for a caus al link
Ab stract
The re la tion ship be tween health care ex pen-
di ture and health out comes is of in ter est to
pol i cy mak ers in the light of steady in creas-
es in health care spend ing for most in dus tri-
alised coun tries. How ev er, es tab lish ing caus-
al re la tion ships is com plex be cause, first ly,
health care ex pen di ture is only one of ma-
ny quan ti ta tive and qual i ta tive fac tors that
con tribute to health out comes, and, sec ond-
ly, mea sure ment of health sta tus is an im-
per fect pro cess. This study re views key find-
ings and method olog i cal ap proach es in this
field and re ports the re sults of our own em-
pir i cal study of coun tries of the Eu ro pean
Union. Our anal y sis ex am ines life ex pect an-
cy and in fant mor tal i ty as the out put of the
health care sys tem, and var i ous life-style, en-
vi ron men tal and oc cu pa tion al fac tors as in-
puts’. Econo met ric anal y ses us ing a fixed ef-
fects mod el are con duct ed on a pan el data
set for the for mer 15 mem bers of the Eu ro-
pean Union over the pe ri od 1980–1995. The
find ings show that in creas es in health ca-
re ex pen di ture are sig nif i cant ly as so ci at ed
with large im prove ments in in fant mor tal i ty
but only marginal ly in re la tion to life ex pect-
an cy. The find ings are gen er al ly con sis tent
with those of sev er al pre vi ous stud ies. Ca-
veats and im prove ments for fu ture re search
are pre sent ed.
Key words
Health care ex pen di ture · Health out comes ·
Ag gre gate data · Mac ro-health
fat (or an i mal prod ucts) were sig nif i cant
[5, 7]. In gen er al, the stud ies re viewed con-
firmed a wors en ing im pact on health out-
comes for ‘neg a tive’ chang es in life-style.
When con sid er ing health ser vice-type ex-
plana to ry vari ables, the num ber of physi-
cians was found to be sig nif i cant in three
stud ies [8, 10, 16], al though the as so ci a tion
was pos i tive for mor tal i ty rates in one of
these [8]. Length of in-pa tient stay and
num ber of beds were found to be sig nif i-
cant in one oth er study [3]. Fi nal ly, stud ies
with less com mon ly used ex plana to ry vari-
ables found, for ex am ple, that liv ing in the
trop ics is as so ci at ed with re duced life ex-
pect an cy [5], and that third-par ty fi nanc-
ing sys tems have an ad verse im pact on in-
fant mor tal i ty [21], con firmed by Elo la et
al. [11] who found that NHS sys tems ha-
ve bet ter in fant mor tal i ty rates for sim i lar
health care ex pen di tures.
The above sum ma ry in di cates some en-
cour ag ing time-se ries re sults, which led
us to re think the case of cross-coun try ti-
me-se ries/cross-sec tion re search for coun-
tries of the EU. We sought to use the best
avail able data and more so phis ti cat ed spec-
i fi ca tion and es ti ma tions of the rel e vant
func tions.
Meth ods for em pir i cal anal y sis
The anal y sis of the re la tion ship be tween
ex pen di tures and health out comes, as out-
lined above, starts from the premise that
health is theout put’ of an ag gre gate pro-
duc tion func tion which ut ilises vari ables
such as health care ex pen di ture, life-sty-
l e , e n v i r o n m e n t a n d o c c u p a t i o n a l f a c t o r s
as the ‘in puts. The as sump tion is that for
rea sons as so ci at ed with di min ish ing re-
turns and the ad verse ef fects of cer tain
vari ables af ter an ini tial pos i tive out come,
the re la tion ship is ex pect ed to be non lin-
ear and non-mono ton ic. For ex am ple, the
ef fects of ris ing in come on health sta tus
are as sumed to be ini tial ly ben e fi cial, but
af ter a cer tain thresh old of in come lev el
which af fects the life style of an in di vid u-
al they may re verse to be come neg a tive,
giv ing rise to a U-shaped func tion. Sim i-
lar con sid er a tions ap ply to bi o log i cal lim-
its and DNA fac tors. More over, in em pir i-
cal re search it is of ten vir tu al ly im pos si ble
to take ac count of the ef fects of la tent vari-
ables, as so ci at ed with im proved nu tri tion
9Eur J Health Econom 1 · 2006
|
Ta ble 1
Sum ma ry of data and re sults for pre vi ous stud ies (PYLL po ten tial years of life lost, DALE dis abil i ty-ad just ed life ex pect an cy,
PPP pur chas ing pow er par i ty, GDP gross do mes tic prod uct, GNP gross na tion al prod uct, GLS gen er alised least squares,
IMR in fant mor tal i ty rate)
Ref er ence De pen dent vari able(s) (out puts) Ex plana to ry vari ables (in puts)
Babazono and
Hill man [3]
Peri na tal mor tal i ty, in fant mor tal i ty,
male life ex pect an cy at birth, fe male
life ex pect an cy at birth, male life
ex pect an cy at 80 years, fe male life ex-
pect an cy at 80 years.
To tal per cap i ta health care spend ing; pub lic per cap i ta health care spend ing (PPP);
in-pa tient beds per 1,000 pop u la tion; ad mis sions per 100 pop u la tion; av er age length
of in-pa tient stay; num ber of physi cians per 1,000 pop u la tion; phy si cian con tacts per
cap i ta; phar ma ceu ti cal ex pen di ture per cap i ta; non-health care spend ing per cap i ta
(PPP); per cent age of pop u la tion aged over 65 years.
Bar low and
Vis sand jee [5]
Life ex pect an cy at birth (males,
fe males and com bined).
Hy po thet i cal max i mum fer til i ty; to tal fer til i ty rate; dai ly in take of an i mal prod ucts;
ac cess to safe wa ter; per cap i ta health ex pen di ture (PPP and ex change rates); per
cap i ta GDP (PPP and ex change rates); pro por tion of adult pop u la tion who are lit er-
ate, ur ban pop u la tion as a pro por tion of to tal pop u la tion; pro por tion of pop u la tion
liv ing in trop ics.
Berg er and Mess er
[7]
Mor tal i ty rate per 1,000 pop u la tion. GDP, health ex pen di ture per cap i ta in U.S.$1990; pop u la tion aged over 65 years; to bac co
con sump tion, al co hol con sump tion; fat con sump tion; fe male labour force par tic i pa tion
rate; pro por tion of pop u la tion aged over 25 years with post-sec ond ary ed u ca tion; Gini
co ef fi cient; pro por tion of to tal health ex pen di tures that are pub licly fi nanced; pro por-
tion of pop u la tion el i gi ble for in-pa tient care ben e fits un der a pub lic health scheme; pro-
por tion of pop u la tion el i gi ble for am bu la to ry care ben e fits un der a pub lic scheme.
Cochrane et al. [8] Age-spe cif ic mor tal i ty rates (ma ter-
nal, peri na tal, in fant, 1-4, 5-14, 15-24,
25-34, 35-44, 45-54, 55-64 years per
10,000 pop u la tion).
Health care (physi cians, nurs es, acute hos pi tal beds, pae da tri cians, mid wives, GNP
spent on health care); di e tary con sump tion; cig a rette con sump tion per cap i ta per an-
num; al co hol con sump tion in litres per cap i ta per an num; calo ries per cap i ta per day;
pro tein per cap i ta per day; fat in take per cap i ta per day; sug ar per cap i ta per day;
de mo graph ic and eco nom ic (pop u la tion den si ty, GNP per cap i ta; ed u ca tion in dex;
in ter ven tion in dex=per cent age of health ex pen di ture cov ered by pub lic ex pen di-
ture).
Crémieux et al. [9] Gen der-spe cif ic in fant mor tal i ty;
gen der-spe cif ic life ex pect an cy at bir-
th and at age 65 years.
Pub lic drug spend ing; pri vate drug spend ing; non-drug health care spend ing; per
cap i ta in come; pop u la tion den si ty; pov er ty; al co hol bev er ages spend ing; gen der-spe-
cif ic to bac co prod ucts spend ing; food and non-
al co hol ic bev er ages spend ing.
Crémieux et al. [10] Gen der-spe cif ic in fant mor tal i ty; gen-
der-spe cif ic life ex pect an cy.
To tal health care spend ing (pri vate and pub lic); per cap i ta physi cians, per cap i ta in-
come; den si ty (pop u la tion/area); ed u ca tion lev el; pov er ty; al co hol use; to bac co use;
nu tri tion al data (meat and fat).
Elo la et al. [11] In fant mor tal i ty rate, PYLL fe males,
PYLL males, life ex pect an cy males, life
ex pect an cy, fe males.
GDP per cap i ta (U.S.$); health care ex pen di ture per cap i ta (U.S.$); pro por tion of pop-
u la tion cov ered by health care sys tem; pub lic health ex pen di tures as pro por tion of
to tal health ex pen di ture; Gini co ef fi cient..
10
|
Eur J Health Econom 1 · 2006
Original Papers
Coun tries stud ied and mod el de scrip tion Prin ci pal re sults
21 OECD coun tries. Aus tralia, Lux em bourg and Turkey
were ex clud ed due to mis sing data. Data are for 1988.
Mod el: mul ti ple lin ear re gres sion us ing step wise
anal y sis (due to small sam ple size).
Num ber of beds and non-health care spend ing are sig nif i cant for peri na tal mor tal i ty
(elas tic i ties −0.52 and −0.48, re spec tive ly) and in fant mor tal i ty (elas tic i ties −0.55 and
−0.35, re spec tive ly); length of stay is sig nif i cant for male life ex pect an cy at birth (elas tic i-
ty=0.6); length of stay (elas tic i ty=0.59) and pub lic health-care spend ing (elas tic i ty=0.38)
are sig nif i cant for fe male life ex pect an cy at birth; non-health care spend ing is sig nif i cant for
male (elas tic i ty=0.5) and fe male (elas tic i ty=0.73) life ex pect an cy. Con clu sion: only fe male
life ex pect an cy at birth is af fect ed by health care ex pen di ture.
76 and 77 de vel oped and de vel op ing coun tries. Data are
for 1990. Mod el: mul ti var i ate re gres sion anal y sis us ing
five-equa tion mod el.
Per cap i ta in come and lit er a cy are strong pre dic tors of life ex pect an cy, their in flu ence ob-
served on prox i mal de ter mi nants (fer til i ty, nu tri tion and wa ter). Health ex pen di ture does
not im pact on life ex pect an cy. Per cap i ta con sump tion of an i mal prod ucts has an in vert-
ed-U re la tion ship with life ex pect an cy, low er fer til i ty is as so ci at ed with ma jor gains in life
ex pect an cy, lo cat ed in the trop ics is as so ci at ed with re duc tions in life ex pect an cy
20 OECD coun tries 1960-1992. Mod el: re gres sion anal y-
sis us ing pan el data and cor rect ed stan dard er rors.
Five mod els are pre sent ed. Sig nif i cant vari ables (co ef fi cients) in most ex ten sive mod el are:
health ex pen di ture (−0.1282), pop u la tion aged over 65 years (0.3334), to bac co (0.1231),
al co hol (0.0477), fat (0.0126), fe male labour force (0.1226); Gini co ef fi cient (−0.096), pro-
por tion of pop u la tion el i gi ble for in-pa tient care ben e fits un der a pub lic health scheme
(0.0821), pro por tion of pop u la tion el i gi ble for am bu la to ry care ben e fits un der a pub lic
scheme (−0.0224).
18 de vel oped coun tries. Data used were for 1970, 1969
or 1971. Mod el: re gres sion anal y ses of mor tal i ty rates on
sev en vari ables found to have the great est ex plana to ry
pow er.
Sev en in put vari ables pro vide the most ex plana to ry pow er: physi cians (pos i tive as so ci a tion
with ma ter nal, peri na tal, in fant and age group 15–24 age group mor tal i ty), GNP (neg a tive
as so ci a tion in most mor tal i ty rates), cigarettes (pos i tive as so ci a tions for all mor tal i ty
rates); al co hol (most ly pos i tive as so ci a tions but neg a tive as so ci a tions for old er age groups);
pop u la tion den si ty (pos i tive as so ci a tion for all but one mor tal i ty rate), in ter ven tion in dex
(most ly neg a tive as so ci a tions); sug ar con sump tion (neg a tive as so ci a tions with all mor tal i ty
rates). Mod el ex plains be tween 42% (5–14 age group) and 97% (in fant mor tal i ty) of vari a-
tion in mor tal i ty rates. Re sults have some anoma lies such as in creas ing physi cians as so ci at-
ed with high er mor tal i ty rates, sug ar in take re duces mor tal i ty rates etc.
Cana di an provinces over the pe ri od 1975–1998. Au-
thors state that data are ho mo ge ne ous in com par i son
with those de rived from in ter na tion al data sets. Mod el:
cross-sec tion al time-se ries GLS for pan el data with cor-
rec tion for AR (1) au to cor re la tion with in pan els and het-
eroskedas tic i ty across pan els. Cana di an provinces are
equal ly weight ed.
Pub lic drug spend ing per cap i ta and pri vate drug spend ing per cap i ta are sig nif i cant for all
health out comes (elas tic i ties: e.g. −0.108 for male in fant mor tal i ty, −0.0.143 for fe male in fant
mor tal i ty, 0.001 for male life ex pect an cy and 0.009 for fe male life ex pect an cy). To tal non-drug
health care spend ing is sig nif i cant for male in fant mor tal i ty, (−0.51), male life ex pect an cy at
birth (0.017) and male life ex pect an cy at age 65 (−0.051). Oth er sig nif i cant vari ables (with
ex pect ed signs) in clude spend ing on al co hol and spend ing on to bac co, spend ing on food
and non-al co hol ic bev er ages, GDP per cap i ta (not for in fant mor tal i ty), den si ty. Sig nif i cant
re gion al vari a tions also ex ist be tween provinces (high er pri vate drug spend ing=bet ter health
out comes than pub lic drug spend ing). If provinces in creased drug spend ing to high est lev els,
584 few er in fant deaths would re sult and over 6 months life ex pect an cy at birth.
Cana di an provinces over the pe ri od 1978-1992. Au thors
state that data are ho mo ge ne ous in com par i son with
those de rived from in ter na tion al data sets. Mod el: ag gre-
gate pro duc tion func tion us ing GLS and provin cial fixed
ef fects. Anal y ses giv en in orig i nal val ues and logs.
Health ex pen di ture is sig nif i cant for all out comes (elas tic i ties: −0.4 for male in fant mor tal-
i ty, −0.6 for fe male in fant mor tal i ty, 0.05 for male life ex pect an cy and 0.024 for fe male life
ex pect an cy). Num ber of physi cians is also sig nif i cant in im prov ing all out comes. Oth er
sig nif i cant vari ables (with ex pect ed signs) are: al co hol con sump tion and per cent age of
smok ers, den si ty=neg a tive im pact fe male life ex pect an cy, pov er ty=neg a tive im pact on
in fant mor tal i ty, meat=pos i tive im pact on fe male life ex pect an cy, in creased fat=neg a tive
im pact on all health out comes ex cept fe male life ex pect an cy, high er in come=high er life
ex pectan cies but not low er in fant mor tal i ty rates.
17 West ern Eu ro pean coun tries (Por tu gal ex clud ed).
Data are for 1990 or 1991, or most re cent avail able at
the time of the study. Data are mean val ues ac cord ing
to health care sys tem (NHS or so cial se cu ri ty). Mod el:
re gres sion anal y sis us ing dum my vari ables for health
care sys tem type (NHS or so cial se cu ri ty). Re la tion ship
be tween in fant mor tal i ty and health ex pen di ture was
in ves ti gat ed af ter con trol ling for GDP.
So cial se cu ri ty sys tems have sig nif i cant ly high er GDP and per cap i ta health ex pen di tures.
Health care ex pen di ture ex plained 32% of vari abil i ty in PYLL and 37% of life ex pect an cy for
fe males. For GDP the fig ures were 26% and 23%, re spec tive ly. Health care ex pen di ture was
a bet ter pre dic tor of in fant mor tal i ty (R
2
=0.45) than GDP (R
2
=0.38). In fant mor tal i ty rates
would be low er for NHS sys tems at sim i lar lev els of health care ex pen di ture (mag ni tude
11–13%).
11Eur J Health Econom 1 · 2006
|
Ta ble 1
Sum ma ry of data and re sults for pre vi ous stud ies (PYLL po ten tial years of life lost, DALE dis abil i ty-ad just ed life ex pect an cy,
PPP pur chas ing pow er par i ty, GDP gross do mes tic prod uct, GNP gross na tion al prod uct, GLS gen er alised least squares,
IMR in fant mor tal i ty rate)
Ref er ence De pen dent vari able(s) (out puts) Ex plana to ry vari ables (in puts)
Grubaugh and
Rex ford [16]
In fant mor tal i ty (health ex pen di ture
also as sessed but not sum marised
here).
Num ber of physi cians per cap i ta; GDP; pop u la tion den si ty; real ed u ca tion ex pen di-
tures per cap i ta; fe male labour force par tic i pa tion rate; per cap i ta real ex pen di tures
of al co hol; per cap i ta real ex pen di ture on to bac co; time trend (for tech nol o gy ef fect);
coun try-spe cif ic non-sys tem; health care sys tem dum my vari able.
Hi tiris and Pos net
[17]
Health ex pen di ture, crude mor tal i ty
rates
GDP; pro por tion of pop u la tion aged over 65 years; per cap i ta health
ex pen di ture.
Leu [21] Age and sex-spe cif ic mor tal i ty rates of
adults; sex-spe cif ic post-neo na tal mor-
tal i ty (2nd to 12th months af ter birth).
GDP per cap i ta; health ex pen di ture; num ber of physi cians (lagged 10 years) and beds
(lagged 10 years); ed u ca tion, ur ban i sa tion; con sump tion of al co hol and to bac co;
pub lic fi nanc ing of med i cal ser vices; NHS fi nanc ing of med i cal ser vices; pro por tion of
pop u la tion aged un der 15 years; di rect democ ra cy.
Licht en berg [22] Life ex pect an cy at birth. Per cap i ta health care ex pen di ture (pri vate and pub lic); med i cal in no va tion (new
drugs and phar ma ceu ti cal R&D).
Mill er and Frech [27] DALE at birth and at age 60 (1998–
1999); life ex pect an cy at birth and at
ages 40 and 60 years (1997–1999);
PYLL for cir cu la to ry dis ease, for can cer,
and for res pi ra to ry dis ease; cause-spe-
cif ic mor tal i ty rates at par tic u lar ages:
35–54, 55–64, 65–74, and 75 years
(1994–1996).
FE MALE, in di ca tor vari able for a fe male out comes mea sure; GDP PC, gross do mes tic
prod uct per cap i ta in PPP; PHPC phar ma ceu ti cal ex pen di tures per cap i ta in PPP;
HEPC oth er health ex pen di tures per cap i ta in PPP; SMOKE, if fe male=1, the per cent-
age of fe males aged 15 years or over who smoke; if fe male=0, the per cent age of ma-
les aged 15 years or over who smoke; AL CO HOL con sump tion per cap i ta; AL CO HOL
×FE MALE, AL CO HOL in ter ac tion with FE MALE; OBE SI TY, pro por tion of fe males with
high body mass in dex.
Or [30] PYLL per 100,000 per sons, aged up to
69 years; all caus es ex cept sui cides.
To tal health ex pen di ture per cap i ta (PPP); pro por tion of pub lic ex pen di ture in to tal
health ex pen di ture; GDP per cap i ta (PPP); pro por tion of white-col lar work ers in to tal
work force; NOx emis sions per cap i ta; al co hol con sump tion; to bac co con sump tion
ex pen di ture per cap i ta (PPP); fat but ter con sump tion per cap i ta; sug ar con sump tion
per cap i ta.
Robali no et al. [32] In fant mor tal i ty ra tio. GDP; de cen tral i sa tion co ef fi cient; struc tur al in di ca tors (pol i tics rights,
cor rup tion, eth nic i ty); coun try ef fect dum my.
Shaw et al. [34] Life ex pect an cy at dif fer ent ages (40,
60, 65 years) for men and wom en in
1997.
Gen der; age; GDP; phar ma exp.; health exp.; be hav iour vari ables (to bac co, but ter and
veg eta bles con sump tion); pol lu tion proxy
Wolfe and Gabay [43] Gen der-spe cif ic life ex pect an cy at
birth and at age 60 years; in fant mor-
tal i ty; pre na tal mor tal i ty rate; med i cal
ex pen di ture.
Med i cal ex pen di ture; pop u la tion aged over 65 years, but ter con sump tion; road ac ci-
dents; liv er cir rho sis (male and fe male); to bac co con sump tion;
em ploy ment in safe and risky in dus tries.
(continued)
12
|
Eur J Health Econom 1 · 2006
Original Papers
Coun tries stud ied and mod el de scrip tion Prin ci pal re sults
Pan el data for 12 OECD coun tries (U.S. ex clud ed). Data
are for 1960–1987. Mod el: mul ti ple re gres sion anal y sis.
Ac tu al and pre dict ed per for mance of the U.S. also anal-
y sed us ing the se lect ed pan el and pub lished data for
the U.S.
Sig nif i cant (co ef fi cient) vari ables for in fant mor tal i ty are: Num ber of physi cians (−0.302),
GDP (−0.386), al co hol (0.099), to bac co (0.145), time trend (−0.145). The U.S. in fant mor tal-
i ty rate (pre dict ed) was 17.2%, while the ac tu al val ue was 12.8% if the U.S. pos sessed the
health care sys tem and un ob serv able non-sys tem struc ture of the typ i cal OECD coun try.
20 OECD coun tries over 28 years (1960–1987). Mod el:
re gres sion anal y ses on three mod els us ing lin ear and
log-lin ear form. Use of coun try-spe cif ic shift dum mies
based on a pooled sam ple of cross-sec tion and time-
se ries data. PPP and ex change rates used.
Mod el 1 con firms the strong link be tween health ex pen di ture and GDP (in come elas tic i ty
of health spend ing=1.026). Mod el 2 shows that pro por tion of age 65 above is sig nif i cant in
ex plain ing health ex pen di ture (elas tic i ty=0.55). Mod el 3 shows that health ex pen di ture has
a neg a tive im pact on mor tal i ty (elas tic i ty low at −0.08), and both pop u la tion aged over 65
years (elas tic i ty=0.350) and GDP (elas tic i ty=0.087) have a pos i tive in flu ence on mor tal i ty
rates. For a giv en lev el of health ex pen di ture and GDP the UK has sig nif i cant ly high er mor-
tal i ty rates in the sam ple used.
19 OECD coun tries (not Lux em bourg, Ice land, Japan,
Por tu gal and Turkey). Data are for 1974. Mod el: re gres-
sion anal y ses us ing some lagged vari ables.
Only re sults for health out comes as de pen dent vari ables giv en here: Vari a tions in adult mor-
tal i ty rates could not be ex plained by mod el. For post-neo na tal mor tal i ty (boys, girls), per
cap i ta GDP (−0.56, −0.71), ed u ca tion (−1.3, −0.62) and pub lic spend ing (−0.09, −0.06) were
sig nif i cant ex plana to ry vari ables. Health ex pen di ture was not sig nif i cant when in come is
con trolled for. Third par ty fi nanc ing sys tems may have an ad verse in flu ence on post-neo na-
tal mor tal i ty rates due to re duced up take.
U.S. over the pe ri od 1960–1997. Mod el: ag gre gate
pro duc tion func tion in cor po rat ing the ge o met ric lag
mod el’ and log a rithms, cor rec tions for se ri al cor re la-
tions.
10% rise in life ex pect an cy from 69.7 to 76.5 years. In creased health ex pen di ture and drug
ap provals ex plain about 100% of ob served long-run in crease in life ex pect an cy. Cost of
med i cal care per life year gained=$11,000; $1,345 for phar ma ceu ti cal R&D. New drugs are
more cost-ef fec tive in in creas ing life ex pect an cy
18 OECD coun tries. Mod el: cross-sec tion re gres sion
anal y sis with lagged vari ables in log-log.
Es ti ma tions with the mod el with life ex pect an cy at birth are non-sig nif i cant (ex cept con-
stant and obe si ty). For all of the oth er mod els with life ex pect an cy at ages 40, at 60, with
DALE at birth and DALE at 60 years: phar ma ceu ti cal ex pen di ture co ef fi cient is al ways sig nif i-
cant and the oth er health ex pen di tures nev er. Ef fects are high er for wom en (elas tic i ty from
0.02 for DALE at birth to 0.09 for DALE at 60). With PYLL and mor tal i ty, re sults are dif fer ent
ac cord ing to the con sid ered pa thol o gy. Obe si ty has also large ef fects. GDP and oth er health
care ex pen di tures are non- sig nif i cant in all mod els ex cept for can cer and res pi ra to ry mor tal-
i ty mod els. Meth o dol o gy seems non-ro bust (prob lems of collinear i ty stressed by au thors),
prob a bly phar ma ceu ti cal catch most of GDP and oth er HC ex pen di ture ef fects.
21 OECD coun tries 1970–1992. Mod el: re gres sion
anal y sis with 483 ob ser va tions us ing pan el data.
Health ex pen di ture is sta tis ti cal ly sig nif i cant on health for wom en, in terms of pre ma ture
death ap prox i mate by PYLL (−0.18 in log), but non-sig nif i cant for men. If GDP re moved
from es ti ma tion (high collinear i ty), both large ly sig nif i cant, and still more for wom en. Pro-
por tion of pub lic health ex pen di ture is sig nif i cant for both men and wom en (−0.17 and
0.18 re spec tive ly). Ma jor con tri bu tion to de crease in pre ma ture mor tal i ty is (re spec tive ly
for wom en and men): (a) Pro por tion of white-col lar (cap tur ing ed u ca tion and work): −0.80
(w), −0.75 (m); (b) GDP per cap i ta PPP: −0.34 (w), −0.44 (m); (c) al co hol: +0.20 (w) and +0.16
(m); (d) pro por tion of pub exp. In to tal health ex pen di ture: −0.17 for both.
67 coun tries (LDC and OECD) 1970–1995. Mod el:
re gres sion anal y sis us ing pan el data.
Fis cal de cen tral i sa tion leads to de crease in IMR (de creas ing re turn ef fects with the GDP lev-
el). Elas tic i ties are around −0.33 for rich er coun tries (those with more than U.S.$6,000 per
cap i ta). Oth er co ef fi cient are also sig nif i cant.
19 OECD coun tries; 1980, 1985, 1990, 1997 data. Mod el:
cross-sec tion anal y sis with lagged vari ables.
Phar ma ceu ti cal exp. leads to in crease life ex pect an cy at ages 60 and 65 years (elas tic i ties
of 0.028 and 0.031, re spec tive ly). Per cap i ta GDP im por tant pre dic tor of life ex pect an cy at
ages 60 and 65 (elas tic i ties of 0.03 and 0.055 re spec tive ly). Es ti mat ed co ef fi cients of oth er
health care ex pen di tures are non-sig nif i cant
22 OECD coun tries in years 1960, 1970 and 1980.
Data are con vert ed into rates of change. Mod el: lin ear
struc tur al re la tions for si mul ta ne ous mod els (mod el
1=health func tion of med i cal ex pen di ture and life-style;
mod el 2=med i cal ex pen di ture is a func tion of life-style).
In creas es in med i cal ex pen di ture lead to im prove ments in all health out comes. Neg a tive
chang es in life-style lead to neg a tive chang es in health out comes and in crease med i cal
ex pen di tures. The in clu sion of life-style vari ables must oc cur in or der to de ter mine the pos-
i tive (ben e fi cial) link be tween health ex pen di ture and health out comes. In crease in pro por-
tion of pop u la tion aged 65 or over, and high er oc cu pa tion al risk are as so ci at ed with high er
med i cal ex pen di ture.
13Eur J Health Econom 1 · 2006
|
and bet ter hy giene (e.g. from a de cline in
wa ter and food-borne dis eases, im proved
wa ter qual i ty and sew age dis pos able sys-
tems, diet, ex er cise, to bac co and al co hol
con sump tion), which have been sug gest-
ed as the most im por tant in the de ter mi-
na tion of health out comes [25].
Tak ing ac count of these caveats and re-
stric tions, we un der took econo met ric anal-
y sis of three de pen dent vari ables as so ci at-
ed with health out comes: life ex pect an cy
at birth of males (M) and fe males (F), and
in fant mor tal i ty (I). The cho sen ex plana-
to ry vari ables, de ter mined by ex am i na-
tion of the OECD health data set and in
light of the vari ables com mon ly ap plied in
pre vi ous re search (as out lined above), for
each of the three equa tions were: to tal (per
cap i ta) health ex pen di ture (U.S.$ PPP), X;
health ex pen di ture as pro por tion of GDP,
Y; num ber of physi cians (per 10,000 head
of pop u la tion), D; num ber of hos pi tal
beds (per 1,000 head of pop u la tion), B;
in-pa tient ad mis sion rate (per cent age of
pop u la tion per an num), A; av er age in-pa-
tient length-of-stay in hos pi tal (days per
an num), H; pop u la tion cov er age of health
care sys tem (per cent age), C; un em ploy-
ment rate, U; al co hol con sump tion (lit-
res per cap i ta per an num), S; ex pen di ture
on to bac co (U.S.$ PPP per cap i ta per an-
num), T; nu tri tion al char ac ter is tics, such
as fruit con sump tion (ki los per cap i ta per
an num), F; nu tri tion (pro tein, per cap i ta
in take per an num), N; and en vi ron men-
tal pol lu tion, P (sul phur ox ide emis sion,
mea sured in ki los per cap i ta per an num).
It was not pos si ble to ob tain cer tain vari-
ables of in ter est, such as ed u ca tion al at tain-
ment for the EU pop u la tion and ac tu al cig-
a rette con sump tion (as op posed to the uti-
lised ex pen di ture on cigarettes), a point ad-
dressed fur ther in the dis cus sion.
The anal y sis was ap plied to data of the
15 EU coun tries of the pe ri od 1980–1995,
i.e. 16×15=240 ob ser va tions. Vari ables
and data were ob tained from the OECD
Health Database [29]. Data were anal y sed
us ing the econo met ric soft ware pack age
SHAZ AM [42]. The vari ables iden ti fied
above are as sumed to ex ert some in flu ence
on the de pen dent vari ables. How ev er, the-
re may be spe cif ic char ac ter is tics in each
coun try as well as la tent vari ables, non-
spec i fied or di rect ly quan tifi able, which
may ex ert di verg ing ef fects on dif fer ent
coun tries. There fore in ad di tion to the ex-
plana to ry vari ables and the con stant term,
our mod el in cludes a dum my vari able for
each EU mem ber state (with the Unit ed
King dom tak en as the stan dard and rep re-
sent ed by the con stant term). Af ter tak ing
ac count of the ex pect ed non-lin ear i ties
and test ing for the spec i fi ca tion form (us-
ing the box Cox sta tis tics) we have cho sen
the log-lin ear func tion al form.
The mem ber-states of the EU make up
a spe cif ic non-ran dom set and, there fore,
the es ti ma tion con cerns a fixed-ef fects mod-
el sub ject to sto chas tic dis tur bances. Con-
se quent ly we ap plied an es ti ma tion meth-
od that takes ac count of the open ness and
in ter de pen dence of the EU economies
with in the com mon mar ket and cor rects
econo met ric prob lems aris ing from the na-
ture of the data in the sam ple by pos tu lat-
ing that the pooled set of coun try data is
cross-sec tion al ly cor re lat ed and time wise
au tore gres sive [19].
Con sis tent es ti mates are de rived by sub-
ject ing the pooled ob ser va tions to or di-
nary least squares es ti ma tion to cal cu late
the cor re spond ing resid u als. These are
used to trans form the vari ables, re move
the au to cor re la tion and, by ap ply ing gen-
er alised least squares, ob tain asymp tot i cal-
ly ef fi cient es ti mates of the re gres sion co ef-
fi cients and their vari ances. For the es ti ma-
tion the data are trans formed, and there-
fore the usu al good ness of fit sta tis tics are
in ap pro pri ate. We used in stead the R
2
sta-
tis tic be tween ob served and pre dict ed val-
ues of the de pen dent vari able.
Re sults
The pleth o ra of cor re lat ed ex plana to ry vari-
ables led to prob lems of multi-collinear i ty.
How ev er, fol low ing man u al step wise pro-
ce dures we elim i nat ed these prob lems by
re mov ing from the set of the ex plana to ry
vari ables and coun try dum my vari ables
those that pro vid ed es ti mat ed co ef fi cients
of size and sign un ac cept able by con ven-
tion al sta tis ti cal and eco nom ic cri te ria.
The es ti mates of the re sult ing par si mo-
nious mod el are pre sent ed in
. Ta ble 2.
The re sults sug gest the fol low ing spe cif-
ic points:
F Male life ex pect an cy: Health ex pen di-
ture, +X, num ber of physi cians, +D,
nu tri tion, +N, and pol lu tion, P, are
sig nif i cant de ter mi nants of male life
ex pect an cy. But there is a sta tis ti cal-
ly sig nif i cant lev el of het ero ge ne ity
be tween coun tries with Swe den the
top per former with 74.8 years, and
Por tu gal and Fin land the worst with
70.9 years, with an EU av er age of
72.7 years in 1990.
F Fe male life ex pect an cy: Health ex pen-
di ture, +X, and num ber of physi cians,
+D, are the sig nif i cant de ter mi nants of
fe male life ex pect an cy. There is mar-
ked het ero ge ne ity be tween coun tries
with France the top per former with
80.9 years, and Ire land the worst with
77.5, with an EU av er age of 79.2 years
in 1990.
F In fant mor tal i ty: Again, health ex-
pen di ture, X, and num ber of physi-
cians, D, are the only sig nif i cant de-
ter mi nants in the re duc tion in in fant
mor tal i ty. The top per former is Swe-
den with an in fant mor tal i ty of 0.6 per
thou sand, the worst is Por tu gal with
1.1, with an EU av er age of 0.8 per thou-
sand in 1990.
. Ta ble 3 pre sents the con tri bu tion of
each ex plana to ry vari able to the out come.
With the ex cep tion of in fant mor tal i-
ty which, dur ing the pe ri od un der re view
has been more than halved by the sig nif i-
cant con tri bu tion of health ex pen di ture, X,
and med i cal care (num ber of physi cians,
D), the pre dom i nant de ter mi nants of both
male and fe male life ex pect an cy are tho-
se con tained in the con stant term, name-
ly the un ac count able salient vari ables and
coun try-spe cif ic char ac ter is tics. There fore
the most im por tant con clu sion reached
by the anal y sis is that health care ex pen di-
ture has made a rel a tive ly mar gin al con tri-
bu tion to the im prove ments in life ex pect-
an cy in the EU coun tries over the pe ri od
of anal y sis. It has added only 2.6 years to
the life ex pect an cy of males and 2.8 years
to that of fe males.
Dis cus sion
As il lus trat ed in this and pre vi ous stud ies,
mea sur ing the im pact of health ex pen di-
ture on health out comes is a com plex and
dif fi cult is sue, which is com mon ly ex am-
ined from ei ther a mi cro- or mac ro-per-
14
|
Eur J Health Econom 1 · 2006
Original Papers
Ta ble 2
Re sults of the es ti ma tions
Life ex pect an cy, males
co ef fi cient (t ra tio)
Life ex pect an cy, fe males
co ef fi cient (t ra tio)
In fant mor tal i ty
co ef fi cient (t ra tio)
In ter cept 4.048 (219.000) 4.120 (427.160) 4.348 (23.922)
Ex pen di ture 0.022 (8.828) 0.022 (17.081) −0.497 (23.466)
Physi cians 0.029 (5.533) 0.034 (11.960) −0.380 (8.412)
Nu tri tion 0.006 (1.992)
Pol lu tion −0.007 (5.106)
Aus tria −0.033 (5.346) −0.167 (3.489) 0.148 (2.566)
Bel gium −0.330 (8.230) −0.255 (11.723) 0.315 (6.367)
Ger many −0.357 (10.902) −0.298 (14187) 0.296 (3.832)
Den mark −0.172 (2.731) −0.260 (4.030)
France −0.258 (8.634)
Fin land −0.255 (5.629) −0.342 (4.629)
Greece 0.169 (2.486)
Italy 0.108 (3.613)
Ire land −0.145 (3.979) −0.250 (6.863)
Lux em bourg −0.305 (11.693) −0.170 (7.706)
Por tu gal −0.382 (9.103) −0.190 (7.735) 0.280 (3.399)
R
2
Buse 0.763 0.733 0.720
R
2
ob served/
pre dict ed
1.000 1.000 0.726
Ta ble 3
Con tribut ing fac tors to health out comes (%) (from: es ti mat ed co ef fi cients in
Table 2)
Male life
ex pect an cy
Fe male life
ex pect an cy
In fant mor tal i ty
% years % years % rate
Health ex pen di ture 3.53 2.6 3.46 2.8 −78.8 −0.63
No. of physi cians 2.14 1.6 2.46 1.9 −27.8 −0.22
Nu tri tion 0.74 0.5
Pol lu tion −0.60 −0.4
Con stant Term 94.19 68.4 94.08 74.5 6.6 1.65
EU av er age 72.7 79.2 0.8
spec tive. Our ap proach has fol lowed the
lat ter, con sid er ing health as the out put of
a health care sys tem, with vari a tions be ing
ex plained by an ar ray of health care in puts
in con junc tion with a num ber of life-style
and en vi ron men tal vari ables.
The re sults of our own em pir i cal study
con firm McK e owns [25] con clu sions for
long-run anal y ses, name ly the rel a tive ly
weak im pact of health care on life ex pect-
an cy, and re search in di cat ing the lim it ed
progress of med i cine in im prov ing health
since the 1980s in de vel oped coun tries [12,
35, 36]. Nev er the less, the more sig nif i cant
con tri bu tion of health care ex pen di ture in
im prov ing in fant mor tal i ty is con sis tent
with the opin ion of some health mac ro-
economists [24]. In terms of life ex pect an-
cy im prove ments it is not ed that fe males
have gained marginal ly more than males,
and anal y sis of the OECD health database
con firms that fe males gen er al ly main tain a
4- to 6-year ad van tage over men for the pe-
ri od of anal y sis. The rea sons for this may
be more com plex than can be dealt with in
this study, but in terms of health ex pen di-
ture it is worth not ing that al most all mass
screen ing pro grammes in de vel oped coun-
tries are tar get ted at wom en (for ex am ple,
breast and cer vi cal can cer), ad di tion al ex-
pen di ture is in curred through child bear-
ing-re lat ed en coun ters with health care
sys tems, and be cause in creased ex pen di-
ture is linked with age ing wom en would
be ex pect ed to uti lise more health care be-
cause of their lon gev i ty.
Be low we high light some im por tant is-
sues that af fect the va lid i ty of the re sults
of pre vi ous stud ies as well as our own find-
ings.
Choice of health out comes
A key is sue in stud ies with our ob jec tive
is the weak ro bust ness of avail able mac ro-
eco nom ic in di ca tors that can be used to
ap prox i mate pop u la tion health sta tus. In-
deed life ex pect an cy and mor tal i ty rates,
com mon ly adopt ed by re searchers, can
only par tial ly re flect the health sta tus of
a pop u la tion and it is dif fi cult to iden ti-
fy feed backs and causal i ty links be tween
health ex pen di tures and health out comes,
es pe cial ly for de vel oped coun tries. It was
felt a pri ori that in fant mor tal i ty would be
a more rep re sen ta tive and re li able health
out come than life ex pect an cy as the lat-
ter is more at tributable to fac tors not re lat-
ed to the health care sys tem, where as the
risks as so ci at ed with child birth and life in
the first year of an in fant are re duced by
bet ter health care fa cil i ties and pro ce dures.
This was borne out by our re sults.
In terms of al ter na tive out come mea-
sures, be cause life ex pect an cy and in fant
mor tal i ty are re gard ed as fair ly crude prox-
ies for health sta tus, which are not very sen-
si tive to chang es in health care fi nanc ing
and de liv ery sys tems [18], a num ber of oth-
er mea sures are be ing de vel oped. These in-
clude amongst oth ers the QALY and HYE
as out lined in the In tro duc tion. Linked to
this re search are a num ber of in stru ments
that mea sure health util i ties, in clud ing the
Med i cal Out comes Study Short-Form 36
and Sick ness Im pact Pro file which have
been adapt ed for use in oth er lan guages
and cul tures [1]. Al though one study in-
clud ed in the re view the DALE [27] was
used, it should be not ed that health sta tus
15Eur J Health Econom 1 · 2006
|
mea sure ments are still in the pro cess of de-
vel op ment and not yet avail able for cross-
coun try com par isons and/or tem po ral
anal y sis (due to lim it ed es ti ma tions). The
choice of life ex pect an cy and in fant mor-
tal i ty to re pres ent health out comes by us
and oth er re searchers in this field of stu-
dy is jus ti fi able. Thus, al though life ex pect-
an cy and in fant mor tal i ty have their lim i-
ta tions, oth ers sup port the view that pop u-
la tion-based mea sures of health out come
have a cru cial place in the over all as sess-
ment of health ser vices. They are the ul ti-
mate val ida tors of so ci etal achieve ment in
re spect of health and as such in vest ment
in their study in ap pro pri ate de tail seems
war rant ed [13].
Mod el mis-spec i fi ca tion
An oth er ma jor dif fi cul ty in mod elling
the re la tion ship be tween health ex pen di-
ture and health out comes is in the po ten-
tial for mod el mis-spec i fi ca tion. In deed,
our anal y sis did not cap ture the im pact of
lagged ef fects, which are par tic u lar ly rel e-
vant to life-style vari ables such as cig a rette
smok ing, al co hol con sump tion and pol lu-
tion. Their im pact on health out comes, as
shown in pre vi ous work, may take a num-
ber of years and anal y ses us ing lag ging
and a large pan el of data would un doubt-
ed ly in crease the va lid i ty of the re sults of
fu ture stud ies.
We also ac knowl edge that our cho sen
mod el could be bet ter spec i fied in terms
of the cho sen vari ables. For ex am ple, we
could not ob tain some vari ables of in ter est
from the OECD Health Database, such as
e d u c a t i o n a t t a i n m e n t a n d a c t u a l c o n s u m p -
tion of cigarettes in the EU coun tries. Our
use of ex pen di ture of cigarettes may not
have cap tured the true im pact of smok ing
in quan ti ta tive terms (al though ex pen di-
ture rather than us age was em ployed in a
num ber of stud ies in clud ed in the re view),
and the im pact on health of ed u ca tion is
well es tab lished in Gross man-type stud ies.
Fu ture re search there fore should ex plore
the pos si bil i ty of in clud ing these and oth er
non-health vari ables rel e vant for this kind
of mod elling.
Fur ther more, the re sults of our re view
in di cate that vari ables on health care sys-
t e m o r g a n i s a t i o n a n d f i n a n c i n g c o u l d b e
em ployed fur ther to test the ef fi cien cy of
var i ous sys tems. In this re gard ex plana to ry
vari ables used may in clude cen tralised/de-
cen tralised sys tems, Bis markian/Bev erid-
gian ap proach es, share of pub lic health ex-
pen di ture, share of pri ma ry care ex pen di-
tures, pri ma ry pre ven tion, ed u ca tion al ca-
re ex pen di tures, and the char ac ter is tics of
health care fi nanc ing.
Data qual i ty
Some ear ly em pir i cal test ing of the re-
la tion ship be tween health ex pen di ture
and health out comes start ed with cross-
sec tion cross-coun try es ti ma tions which,
prob a bly for rea sons as so ci at ed with da-
ta het ero ge ne ity and poor qual i ty (e.g. the
OECD start ed col lect ing and pub lish ing
sta tis ti cal se ries on health only in the mid-
dle 1990s) did not pro vide any sup port to
the hy poth e sis of the link be tween health
ex pen di ture and health out comes. In deed,
pre vi ous stud ies have shown that mea sur-
ing health out comes for any coun try will
have its lim i ta tions, and these lim i ta tions
will be fur ther ex ac er bat ed by the like li-
hood that a group of 15 coun tries (such
as the EU over the pe ri od of anal y sis) will
have some vari a tions in the way that each
coun try de fines and records data [1]. As
Mac beth [23] rather pes simisti cal ly points
out, this di ver si ty of def i ni tions ‘in val i-
dates many com par isons.
As in di cat ed above, more re cent data
would help to in crease the re li abil i ty of our
own anal y sis. How ev er, in con sid er ing the
up dat ing of our data pan el we not ed vari-
a tions in def i ni tions for the OECD data
set be tween the ver sion we used and new-
er ver sions, as well as gaps for some vari-
ables/coun tries. In fu ture work we will ex-
plore the pos si bil i ty of us ing oth er sources
to up date and ex tend our dataset. By util-
is ing more re cent and ex tend ed vari ables
in our dataset the pres ent find ings can be
fur ther ex plored and val i dat ed. [For ex am-
ple, the Sum mer and Pre ston 1991 dataset,
the Penn World Data (PWD) is avail able
at: http://bized.ac.uk/dataserv/pennhome.
htm; the Sachs-Warn er dataset is avail able
at: http://www.nuff.ox.ac.uk/Eco nomics/
Growth/datasets.htm; the East er ly-Le-
vine dataset, the Bar ro-Lee dataset and
t h e Wo r l d D e v e l o p m e n t I n d i c a t o r s 2 0 0 1
are avail able at: http://www.world bank.
org/data; the WHO Eu ro pean Health For
All Database (HFA), 2005 edi tion, is avail-
able at: http://www.who.dk.]
De vel oped vs. de vel op ing
coun tries
Our re sults show a mar gin al but pos i tive
ef fect for health ex pen di ture on the ex am-
ined health out comes for de vel oped na-
tions (rep re sent ed by EU), more so for in-
fant mor tal i ty than life ex pect an cy, which
is con sis tent with ev i dence con firm ing di-
min ish ing re turns in the area of health ca-
re in de vel oped coun tries [4, 33]. In con-
trast, small amounts of health ex pen di ture
in de vel op ing coun tries, and even in ter-
me di ate coun tries, would al most cer tain-
ly have a big ger im pact. In this re gard it
would be an in for ma tive and in ter est ing
ex er cise to link these re sults with the re-
search field on growth and health, where
this re la tion ship [6] has been stressed by
many au thors [37, 40] for health ex pen di-
ture on growth. Bau mol [6] showed that
due to pro duc tiv i ty dif fer ences be tween
the ser vices sec tor (‘low pro duc tiv i ty’ ac-
tiv i ties) and the in dus tri al sec tor (‘high
pro duc tiv i ty’ ac tiv i ties) the val ue share of
the for mer will in crease over time, i.e. a
grow ing part of na tion al in come will be
spent (and earned) through the low pro-
duc tiv i ty sec tor, as long as there is a de-
mand for these ac tiv i ties.
This well-known phe nom e non is cal-
led ‘Bau mols dis ease’ and health ser vices
is one of the usu al ap pli ca tions of it. In-
deed, the de mand for health ser vices (de-
rived from the con cept of needs) is re cog-
ni sed as un lim it ed as long as peo ple ex-
pe ri ence (ex og e nous) vari a tions in their
health sta tus. More over, the low pro duc-
tiv i ty char ac ter of the health sec tor is en-
hanced be cause the pro vi sion of health ser-
vices is sub ject to a de creas ing re turns to
scale of 1. It would be an in ter est ing is sue
to ex plore new re search to val i date Bau-
mols dis ease in health care in the light of
the lim it ed im pact of care on glob al health
sta tus in de vel oped coun tries over the past
few decades. As pre vi ous ly out lined, it is
the case that since the 1980s in the de vel-
oped coun tries there have been no cru-
cial in no va tions in health care, i.e. hav-
ing a large im pact on the pop u la tion as a
whole in these coun tries, while health care
ex pen di tures have in creased dra mat i cal ly.
16
|
Eur J Health Econom 1 · 2006
Original Papers
Hence the phe nom e non de scribed by Bau-
mol could ex plain the large de creas ing re-
turn (con verg ing to wards 0) of health ca-
re ex pen di ture on health out comes sin-
ce in creas es in ex pen di ture are large ly tak-
en up among ser vices (work force) grow-
th and the de vel op ment of ex pen sive prod-
ucts which are ap pli ca ble to small sub-pop-
u la tions. In con trast, the dis cov ery of new
vac cines or large ef fect drugs, such as pen-
i cil lin, both pro duc ing huge pos i tive ex ter-
nal i ties, in duced in creas ing re turns based
on prod ucts (not ser vices). One may con-
tem plate that this could be again the case
in the next decades with ge net ics and as so-
ci at ed ther a pies.
In sum ma ry, in stud ies that have used
ag gre gate data to ex plore caus al links be-
tween health out comes and ex plana to ry
vari ables such as health ex pen di ture, those
as so ci at ed with the health care sys tem, en-
vi ron men tal and life-style, there are some
con flict ing re sults and method olog i cal is-
sues that need to be ad dressed, sug gest ing
more work needs to be done in this area
[44]. The above dis cus sion re gard ing meth-
o dol o gy and data sources should, in par tic-
u lar, help to im prove fu ture stud ies.
Con clu sions
This study ex am ined the con tri bu tion
of health care ex pen di ture to health out-
comes. The find ings show that is sues in-
volved here are com pli cat ed be cause, first,
health ex pen di ture is only one of many
quan ti ta tive and qual i ta tive fac tors that
con tribute to health out comes, and, sec-
ond, health out comes are also qual i ta tive
and quan ti ta tive, and only the lat ter may
be as sessed by the avail able sta tis ti cal and
econo met ric tech niques. Tak ing ac count
of these con straints we searched for the
de ter mi nants and their ef fects on three
con ven tion al health out comes, male and
fe male life ex pect an cy at birth and in fant
mor tal i ty, for which quan ti ta tive sta tis ti cal
ob ser va tions are avail able. In the pro cess
we were con front ed with the sev er al but
ex pect ed econo met ric ob sta cles and prob-
lems but ar rived at rea son able re sults with-
in the con straints found in us ing OECD
health data. In terms of the lim i ta tions of
our em pir i cal study we ac knowl edge that
the size of the avail able sam ple did not per-
mit us to test for the pos si ble ex is tence of
any lag struc ture in the ex plana to ry vari-
ables, po ten tial ly rel e vant to ex pen di ture
on cigarettes, al co hol con sump tion and en-
vi ron men tal in flu ences. Giv en these cave-
ats, the find ings of this study lead to the
con clu sion that while health care ex pen di-
tures are among the most im por tant fac-
tors in the low er ing of in fant mor tal i ty,
they make only a mar gin al con tri bu tion
to the im prove ment of male and fe male
life ex pect an cy. Our re sults are broad ly in
line with those of oth er stud ies re viewed
in this con tri bu tion on de vel oped coun-
tries, but there are many caveats that need
to be con sid ered when in ter pret ing the re-
sults of this and oth er stud ies, which we
have at tempt ed to iden ti fy and dis cuss to
the ben e fit of fu ture stud ies in this area of
re search.
Cor re spond ing au thor
John Nixon
Cen tre for Re views and Dis sem i na tion,
Uni ver si ty of York, York, YO10 5DD, UK
e-mail: jn105@york.ac.uk
Ac knowl edge ments
We ex press our thanks to Dr. Theo Hi tiris, for mer-
ly of the De part ment of Eco nomics and Re lat ed
Stud ies, Uni ver si ty of York, for his in put to the
econo met ric anal y ses and con struc tive crit i cisms
in the prep a ra tion of this manuscript.
Con flict of in ter est: No in for ma tion sup plied
Ref er ences
1. An der son GF, Poul li er J-P (1999) Health spend-
ing, ac cess, and out comes: trends in in dus tri al ized
coun tries. Health Aff (Mill wood) 18:178–192
2. Ar row K (1963) So cial choice and in di vid u al val ues.
Wi ley: New York
3. Babazono A, Hill man AL (1994) A com par i son of
in ter na tion al health out comes and health care
spend ing. Int J Tech nol As sess Health Care 10:376–
381
4. Bank W (1993) World de vel op ment re port: in vest-
ing in health. Ox ford Uni ver si ty Press: New York
5. Bar low R, Vis sand jee B (1999) De ter mi nants of na-
tion al life ex pect an cy. Can J Dev Stud 20:9–28
6. Bau mol WJ (1967) Mac ro-eco nomics of un bal-
anced growth: the anat o my or ur ban cri sis. Am
Econ Rev 57:415–426
7. Berg er MC, Mess er J (2002) Pub lic fi nanc ing of
health ex pen di tures, in sur ance, and health out-
comes. Appl Econ 34:2105–2113
8. Cochrane AL, St Ledger AS, Moore F (1978) Health
ser vice ‘in put’ and mor tal i ty out put’ in de vel oped
coun tries. J Epi demi ol Com mu ni ty Health 32:200–
205
9. Crémieux P-Y, Mieilleur M-C, Ouel lette P, Petit P,
Zelder P, Potvin K (2005) Pub lic and pri vate phar ma-
ceu ti cal spend ing as de ter mi nants of health out-
comes in Can a da. Health Econ 14:107–116
10. Crémieux P-Y, Ouel lette P, Pi lon C (1999) Health ca-
re spend ing as de ter mi nants of health out comes.
Health Econ 8:627–639
11. Elo la J, Daponte A, Vi cente N (1995) Health in di ca-
tors and the or ga ni za tion of health care sys tems in
West ern Eu rope. Am J Pub lic Health 85:1397–1401
12. Fuchs VR (1986) The health econ o my. Har vard Uni-
ver si ty Press: Cam bridge
13. Goldacre M (1996) Health out comes from a med-
i cal per spec tive. In: Mac beth HM (ed) Health out-
comes, bi o log i cal, so cial and eco nom ic per spec-
tives. Ox ford Uni ver si ty Press: Ox ford
14. Gross man M (1972) On the con cept of health cap-
i tal and the de mand for health. J Po lit i cal Econ
80:223–255
15. Gross man M (2000) The hu man cap i tal mod el. In:
Cu ly er AJ, New house JP (ed) Hand book of health
eco nomics, vol 1A. Else vier: Ams ter dam
16. Grubaugh SG, Rex ford ES (1994) Com par ing the
per for mance of health-care sys tems: an al ter na-
tive ap proach. South Econ J 60:1030–1042
17. Hi tiris T, Poss nett J (1992) The de ter mi nants and ef-
fects of health ex pen di ture in de vel oped coun tries.
J Health Econ 6:173–181
18. Kindig DA (1997) Pur chas ing pop u la tion health:
pay ing for the re sults. Uni ver si ty of Michi gan Press:
Ann Ar bor
19. Kmen ta J (1986) El e ments of econo met rics. Col lier
Macmil lan: Lon don
20. Le Galès C, Buron C, Costet N, Ros man S, Sla ma
G (2002) De vel op ment of a pref er ence-weight ed
health sta tus clas si fi ca tion sys tem in France: the
Health Util i ties In dex. Health Care Man age Sci
5:41–52
21. Leu RE (1986) The pub lic-pri vate mix and in ter na-
tion al health care costs. In: Cu ly er AJ, Jöns son B
(ed) Pub lic and pri vate health ser vices. Black well
Ba sil: Ox ford, pp 41–63
22. Licht en berg F (2000) Sources of U.S. lengevi ty in-
crease, 1960–1997. Cen ter for Eco nom ic Stud ies,
and Ifo In sti tute for Eco nom ic Re search (CE Si fo),
Work ing Pa per Se ries, Mu nich, Ger many, work ing
pa per no 405
23. Mac beth HM (1996) Health out comes, bi o log i cal,
so cial and eco nom ic per spec tives. Ox ford Uni ver si-
ty Press: Ox ford
24. Ma jnoni d’Intig nano B, Ul mann P (1999) The qual-
i ta tive per for mance of the French health care sys-
tem – evo lu tions com pared to Eu rope since 1970.
Health Sci Syst 3:123–141
25. McK e own T (1979) The Role of med i cine: dream,
mi rage or ne mi sis? Ba sil Black well: Ox ford
26. Mehrez A, Gafni A (1989) Qual i ty ad just ed life
years (QALYs) util i ty the o ry and heal thy years
equiv a lent (HYE). Med De cis Mak ing 9:142–149
27. Mill er RD, Frech T (2002) The pro duc tiv i ty of health
care and phar ma ceu ti cals: qual i ty of life, cause. De-
part ment of Eco nomics, UCSB de part men tal work-
ing pa pers, 12-02, 1,-42. Uni ver si ty of Cal i for nia,
San ta Bar bara
28. New house JP (1977) Med i cal-care ex pen di ture: a
cross-na tion al sur vey. J Hum Re sour 12:115–125
29. OECD/CRE DES (2000) OECD health data 2000. A
soft ware pack age for in ter na tion al com par i son of
health care sys tems. OECD/CRE DES: Par is
30. Or Z (2000) De ter mi nants of health out comes in in-
dus tri alised coun tries: a pooled, cross-coun try, ti-
me-se ries anal y sis. OECD: Par is, pp 53–77
17Eur J Health Econom 1 · 2006
|
31. Pi ate c ki C, Ul mann P (1996) The dilem na of the
in di vid u al and the col lec tive: growth and health.
2nd In ter na tion al Con fer ence on Pub lic Eco nomics,
Uni ver si ty of Brest
32. Robali no DA, Os car FP, Al ber tus V (2001) Does fis-
cal de cen tral iza tion im prove health out comes? Ev-
i dence from a cross-coun try anal y sis. World Bank
Pol i cy re search work ing pa per se ries 2565:1–14
33. Sachs JD, Warn er AM (1997) Sources of slow grow-
th in African economies. J Afr Econ 6:335–376
34. Shaw JW, Hor race WC, Vo gel RJ (2002) The pro-
duc tiv i ty of phar ma ceu ti cals in im prov ing health:
an anal y sis of the OECD health data. WUSTL
eco nomics work ing pa per archive, HEW se ries
0206001:1–42
35. Stod dart G (1995) The chal lenge of pro duc ing
health in mod ern economies. Work ing pa per no
46. Cana di an In sti tute for Ad vanced Re search:
Toron to
36. Ul mann P (1998) Health eco nomics: some stylised
facts. Health Sci Sys tem 1.2:309–356
37. Ul mann P (2002) La Prise en compte de la san-
té dans les mod èles de crois san ce: bi lan et per-
spec tives Tak ing into ac count health in eco nom ic
growth mod els: as sess ment and prospects. In: Du-
puis J-M et al. (ed) Poli tiques so ciales et crois san ce
économique, vol 1. L’Har mat tan: Par is, pp 9–30
38. Van Doorslaer E, Wagstaff A, Ble ichrodt H, Ca longe
S, Gerdtham U-G, Gerfin M (1997) In come-re lat ed
in equal i ties in health: some in ter na tion al com par-
isons. J Health Econ 16:93–112
39. Van Our ti T (2003) So cio-eco nom ic in equal i ty in
ill-health amongst the el der ly. Should one use cur-
rent or per ma nent in come? J Health Econ 22:219–
241
40. Zon A van, Muysken J (2001) Health and en dog e-
nous growth. J Health Econ 20:169–185
41. Wagstaff A, Van Doorslaer E, Paci P (1989) Eq ui ty in
the fi nance and de liv ery of health care: some ten ta-
tive cross-coun try com par isons. Ox ford Rev Econ
Pol i cy 5:89–113
42. White KJ (1993) Shaz am econo met rics com put er
pro gram: ver sion 7.0 user’s ref er ence man u al. Mc-
Graw-Hill: New York
43. Wolfe BL, Gabay M (1987) Health sta tus and med i-
cal ex pen di tures: more ev i dence of a link. Soc Sci
Med 25:883–888
44. Zweifel P, Brey er F (1997) Health eco nomics. Ox-
ford Uni ver si ty Press: Ox ford
18
|
Eur J Health Econom 1 · 2006
The corresponding author is asked to
declare any existing connection between
the author(s) and (a) the supplier of any
product mentioned in this article or (b) any
supplier of a competing product.
Manuscript Preparation and Submission
Send an electronic file of your manuscript to:
Professor Dr. Wolfgang Greiner,
Universität Bielefeld, Fakultät für Ge sund-
heits wissenschaften, Gesundheitsökonomie
und Gesundheitsmanagement (AG5) Uni-
versität Bielefeld, Postfach 100131,
33501 Bielefeld, Germany, Tel +49-521-106
6989, Fax +49-521-106 8054, E-mail: EJHE@
uni-bielefeld.de. It should be submitted in
duplicate, typed on one side of the paper
only, double-spaced, with wide margins. The
text should be as concise as possible. Origi-
nal papers may contain 14 typewritten pag-
es subtracting space for 12 illustrations. Re-
view articles should consist of not more than
20 typewritten pages. This will include space
for 16 illustrations. Form and content of the
manuscript should be carefully checked to
preclude the need for corrections in proof.
Please mark in the margin of the manuscript
the desired position of figures and tables.
Language. Papers should preferably be sub-
mitted in English. It is also possible to submit
manuscripts in French, Spanish and Italian.
Please note that this might delay the review-
process.
The title page should comprise: title of pa-
per, first name(s) and surname of author(s),
institute affiliations (given as footnotes) and
any other footnotes referring to the title (in-
dicated by asterisks), running title (not more
than 72 typewriter characters, incl. spaces),
address to which proofs should be sent.
Abstract. Each paper should be preceded by
an abstract.
Key words. Immediately following the ab-
stract not more than five key words should
be supplied for subject indexing. Key words
should be taken from the Index Medicus
(Medical Subject Headings) or, failing this, be
composed of the same principles.
Footnotes. Other than those referring to the
title heading, footnotes should be numbered
consecutively.
Figures. The number and size of the illustra-
tions must be kept to the minimum required
clarification of the text. All facial photographs
must be accompanied by an appropriate
release form. Explanations of figures fur-
nished as legends should not be repeated in
the text. Numerical data given in graphs or
tables must not duplicate each other. All fig-
ures, whether photographs, graphs or dia-
grams, should be numbered consecutively
throughout and submitted on separate
sheets. Where there are several figures, they
should be grouped to fit on a single page.
Line Drawings. Please submit good-quality
glossy prints in the desired final size. The in-
scriptions should be clearly legible.
Half-Tone Illustrations. Please submit well-
contrasted regular photographic prints (not
Polaroid), trimmed at right angles, preferably
in the desired final size.
Captions. Each figure should be briefly and
clearly described. Remarks like: ’For explana-
tion see text’ should be avoided. Captions
are a part of the text and should be append-
ed to it on a separate sheet.
Bibliography. Include only works referred to
in the text. References should be cited as
follows: Journal papers: name(s) and initial(s)
of author(s) (no periods), year of publication
(in parentheses), complete title of paper
(followed by a period), name of journal as
abbreviated in Index Medicus (no periods),
volume number, first and last page number
(separated by a hyphen with no period at
the end); Books: name(s) and initial(s) of
author(s) (no periods), year of publication (in
parentheses), full title, edition, publishers,
place, pages.
Copyright. Submission of a manuscript im-
plies: that the work described has not been
published be fore (except in the form of an
abstract or as a published lecture, review, or
thesis); that it is not under consideration
elsewhere; that its publication has been ap-
proved by all co authors, if any, as well as by
the respon sible authorities at the institute
where the work has been carried out; that, if
and when the manuscript is accepted for
publication, the authors agree to automatic
transfer of the copyright to the publisher;
that the manu script will not be published
elsewhere in any language without the writ-
ten consent of the copyright holders; that
written permission of the copyright holder is
obtained by the authors for material used
from other copy righted sources; and that
any costs asso ciated with obtaining this per-
mission are the authors’ responsibility. Trans-
fer of copyright to Springer-Verlag becomes
effective if and when the article is accepted
for publication. The copyright covers the ex-
clusive right (for U.S. government employ-
ees: to the extent transferable) to reproduce
and distribute the article, including reprints,
translations, pho to graphic reproductions,
microform, elec tro nic form (off line, online) or
other re pro duc tions of similar nature.
© Springer Medizin Verlag
Instructions for authors
... Most studies focus only on a narrow selection, whilst some consider the linkages between LEAB and the three pillars of sustainability (economic, social and environmental), e.g., [8,[13][14][15][16]. With some notable exceptions, however, there is a relative dearth of literature analysing LEAB in Europe [8,17,18], where most applications restrict themselves to a handful of countries, thereby ignoring the relative performance of the European continent as a whole and the heterogeneity of outcomes across individual countries. ...
... where H j is the health output of individual j and X j is a vector of inputs included in the health production function F. According to a number of commentators [13,[15][16][17], this microeconomic framework can be scaled up to the macroeconomic level without losing its theoretical grounding by considering per capita or average data, with numerous examples in the recent literature [9, 13-18, 23, 26]. Moreover, macro data analyses have the advantage that the effect of the drivers of LEAB of the overall population can be obtained, providing valuable insights for policy-making. ...
... The quality of public services is also recognised as a key factor influencing people's health; an aspect that macroeconomic studies usually proxy through health expenditure data. Importantly, however, a cursory glance at the literature reveals that the effect of this variable remains ambiguous [9,13,15,17,18,26,30]. Indeed, the consensus in the above studies is that the expectation that an increase in health expenditure may improve health services and hence health status, is only true if the marginal effect of this increase is greater than the forgone benefits that would have accrued ...
Article
Full-text available
The European continent has one of the longest life expectancies in the world, but still faces a significant challenge to meet the health targets set by the Sustainable Development Goals of the United Nations for 2030. To improve the understanding of the rationale that guides health outcomes in Europe, this study assesses the direction and magnitude effects of the drivers that contribute to explain life expectancy at birth across 30 European countries for the period 2008–2018 at macro-level. For this purpose, an aggregated health production function is used allowing for spatial effects. The results indicate that an increase in the income level, health expenditure, trade openness, education attainment, or urbanisation might lead to an increase in life expectancy at birth, whereas calories intake or quantity of air pollutants have a negative impact on this health indicator. This implies that health policies should look beyond economic factors and focus also on social and environmental drivers. The results also indicate the existence of significant spillover effects, highlighting the need for coordinated European policies that account for the synergies between countries. Finally, a foresight analysis is conducted to obtain projections for 2030 under different socioeconomic pathways. Results reveal significant differences on longevity projections depending on the adoption, or not, of a more sustainable model of human development and provides valuable insight on the need for anticipatory planning measures to make longer life-spans compatible with the maintenance of the welfare state.
... Most studies focus only on a narrow selection, whilst some consider the linkages between LEAB and the three pillars of sustainability (economic, social and environmental), e.g., [8,[13][14][15][16]. With some notable exceptions, however, there is a relative dearth of literature analysing LEAB in Europe [8,17,18], where most applications restrict themselves to a handful of countries, thereby ignoring the relative performance of the European continent as a whole and the heterogeneity of outcomes across individual countries. ...
... where H j is the health output of individual j and X j is a vector of inputs included in the health production function F. According to a number of commentators [13,[15][16][17], this microeconomic framework can be scaled up to the macroeconomic level without losing its theoretical grounding by considering per capita or average data, with numerous examples in the recent literature [9, 13-18, 23, 26]. Moreover, macro data analyses have the advantage that the effect of the drivers of LEAB of the overall population can be obtained, providing valuable insights for policy-making. ...
... The quality of public services is also recognised as a key factor influencing people's health; an aspect that macroeconomic studies usually proxy through health expenditure data. Importantly, however, a cursory glance at the literature reveals that the effect of this variable remains ambiguous [9,13,15,17,18,26,30]. Indeed, the consensus in the above studies is that the expectation that an increase in health expenditure may improve health services and hence health status, is only true if the marginal effect of this increase is greater than the forgone benefits that would have accrued ...
Article
Full-text available
The European continent has one of the longest life expectancies in the world, but still faces a significant challenge to meet the health targets set by the Sustainable Development Goals of the United Nations for 2030. To improve the understanding of the rationale that guides health outcomes in Europe, this study assesses the direction and magnitude effects of the drivers that contribute to explain life expectancy at birth across 30 European countries for the period 2008-2018 at macro-level. For this purpose, an aggregated health production function is used allowing for spatial effects. The results indicate that an increase in the income level, health expenditure, trade openness, education attainment, or urbanisation might lead to an increase in life expectancy at birth, whereas calories intake or quantity of air pollutants have a negative impact on this health indicator. This implies that health policies should look beyond economic factors and focus also on social and environmental drivers. The results also indicate the existence of significant spillover effects, highlighting the need for coordinated European policies that account for the synergies between countries. Finally, a foresight analysis is conducted to obtain projections for 2030 under different socioeconomic pathways. Results reveal significant differences on longevity projections depending on the adoption, or not, of a more sustainable model of human development and provides valuable insight on the need for anticipa-tory planning measures to make longer lifespans compatible with the maintenance of the welfare state.
... Country and regional disparities in life expectancies (LE) have been long attached to income levels, and, more recently, to healthcare expenditure (Cremieux et al., 1999(Cremieux et al., , 2005Elola et al., 1995;Shaw et al., 2002;Nixon & Ulman, 2006;Jain & Yuan, 2020;Martín-Cervantes, Rueda López, & Cruz Rambaud, 2019). However, since LE are achieving very high levels, researchers have turn into looking at the macro determinants behind differences in morbidity measures, specifically on different indicators of healthy life expectancies (Gutierrez-Fisac et al., 2000;Groenewegen et al., 2003;Kondo et al., 2005;Jagger et al., 2008;Liu et al., 2010;Fourweather et al., 2015;Minagawa & Saito, 2018;Laborde et al., 2021). ...
Article
Full-text available
Life expectancy has long been associated with macro-level factors, including health expenditures, but little research has focused on the relationship with morbidity measures. This paper examines the relationship between the expected years lived free of and with chronic conditions (YLFCC and YLCC) at age 50 and macroeconomic and social factors including, for the first time, several indicators of public health expenditure. We calculate YLFCC and YLCC for Spanish regions using the Sullivan method over a long period of time (2006–2019). Spain is a good case study due to two reasons. First, its national health system is decentralized among regional administrations since 2002. Second, the financial crisis of 2008 led to public health cuts in 2010–2014 that each region handled differently. We use fixed-effects models to assess the relationship between changes in macro-level regional indicators (socioeconomic factors, healthcare resources, health behavior and public health expenditures) with YLFCC and YLCC across regions and over time. Results show that socioeconomic levels, public health expenditure, healthcare resources and health behaviors are associated with years lived free of and with chronic conditions when analyzing them independently. However, in the global model including all these dimensions only public health expenditure is associated with both YLFCC and YLCC for men and women, showing that a higher level of expenditures is correlated with more YLFCC and less YLCC. Therefore, regional authorities need to pay special attention to the level of investments on health services, as they are clearly associated with a better quality of living of the middle age and older population.
... Note that the revised table with the re-estimated findings is not included here due to space constraints, but is available upon request. Indeed, both, life expectancy and infant mortality are useful proxies for assessing a population's health since it is widely anticipated that improvements in health inputs would lead to better health outcomes (Jaba et al., 2014;Shaw et al., 2005;Nixon and Ulmann, 2006). The selection of these two outcomes (life expectancy & infant mortality) was prompted by a paucity of data on other health inputs in a significant number GCC Countries. ...
Article
The paper aims to analyse the long run and short run determinants of health in Gulf Corporation Council (GCC) Countries. Based on the ARDL method, our analysis showed that none of the GCC countries have shown any significant positive effect of resources rents on life expectancy, and that resource curse exists in all GCC countries in the short run and long run. In light of institutional quality, we found that in the short run, corruption had a significant negative effect on life expectancy in all of GCC countries, except for Kuwait and Qatar, whereas in the long run GCC countries showed a significant negative effect, except Kuwait. Moreover, democracy had a statistically insignificant effect on life expectancy for almost all GCC countries in the short run as well as in long run, except for Kuwait and Qatar. To check for robustness in the long run, our analysis showed that the ARDL results are confirmed by the results of different methods, namely: the fully modified ordinary least square (FMOLS), dynamic ordinary least square (DOLS), and the canonical cointegrating regression (CCR) methods. The findings of this paper show a strong evidence of the moderation effect of corruption and democracy on the relationship between life expectancy and resources rent. The quality of institutions represented by low corruption and high democracy rates will be sufficient enough to reverse the negative relationship between resources rent and life expectancy.
... Life expectancy has improved globally [1], and the average life expectancy in many industrialised countries is expected to be 85 years or more by 2030 [2,3]. A prolonged lifetime is attributed to social factors, such as sociodemographic status [1], gross domestic product per capita [4,5], healthcare expenditure and universal health coverage [6,7] and policies and legislation [8]. Meanwhile, the joint impact of selected modifiable healthy lifestyle factors, such as moderate physical activity [9], healthy body-mass index (BMI) [10], nonsmoking status [11,12], moderate alcohol intake [13,14] and appropriate sleep duration [15,16] has been associated with an increased lifespan in industrialised countries [17][18][19][20][21][22][23]. ...
Article
Objective: this study explored whether the modification of selected lifestyles is likely to increase life expectancy from middle age onwards, regardless of the presence of major comorbidities. Methods: we examined a prospective cohort of 20,373 men and 26,247 women aged 40-80 years. Eight modifiable lifestyle factors were assessed: consumption of fruit, fish and milk, walking and/or sports participation, body-mass index, smoking status, alcohol consumption and sleep duration. Modifiable healthy lifestyle factors scored one point each, for a maximum of eight points. The impact of modifiable healthy lifestyle adoption on lifetime gain during the ages of 40-102 years was analysed. Findings: during the median 21 years of follow-up, 8,966 individuals (3,683 men and 5,283 women) died. Life expectancy at 40 years (95% confidence intervals) for 7-8 health lifestyle points was 46.8 (45.6-48.1) and 51.3 (50.0-52.6) years for men and women, respectively. The potential impact of modifiable healthy lifestyle adoption on lifetime gain persisted over the age of 80 years or more, in individuals with ≥5 factors (P < 0.001), particularly older men. The benefits were more pronounced among patients with major comorbidities, such as cardiovascular disease, cancer, hypertension, diabetes, kidney disease and those with multimorbidity throughout all age categories. Conclusion: adopting modifiable healthy lifestyles was associated with lifetime gain, even in individuals aged 80 years or more, regardless of the presence of any major comorbidities in each life stage since middle age. The findings imply the importance of improving the one's lifestyle for an increased lifespan, even among older patients and/or those with multimorbidity.
... A number of studies have investigated the relationships between health expenditure and health outcomes using different measures of health status such as life expectancy and infant and child mortality (Baltagi & Moscone, 2010;Barenberg et al., 2017;Gallet & Doucouliagos, 2017;Linden & Ray, 2017;Nixon & Ulmann, 2006). For developing countries such as those in Africa, infant mortality and under-5 mortality rates are considered to be a flash indicator of the health conditions because data on these indicators are more reliable than other indicators such as life expectancy and maternal mortality (Bokhari et al., 2007;Mishra & Newhouse, 2009;Obrizan & Wehby, 2018). ...
Article
This paper examines the effect of public and private health spending on health outcomes measured by infant and child mortality rate using a panel of 37 African countries over the period 1995–2018. Based on both system generalized method of moments and least squared dummy variable corrected bias estimators, we find that the level of public health expenditures has statistically significant positive effects on mortality reduction. However, private health spending has a statistically significant positive effect on infant and child mortality rates when it interacts with public health spending. This means that the health effect of private health expenditure is strengthened by public health spending. Conditioning on the level of private health expenditure, the total effect of public health expenditure on child mortality reduction is slightly higher in absolute value than that of private health expenditure. The main policy implication that emerges from our results is the need to increase public health expenditure and support the effectiveness of private health expenditure. This additional fund should be allocated to some specific expenditure categories which have a very direct link with child and infant mortality rates.
Article
Resumo Concebido pela Constituição de 1988, o Sistema Único de Saúde (SUS) vislumbra a descentralização fiscal e a distribuição equitativa de fundos para estados e municípios. Por meio de uma Análise Exploratória de Dados Espaciais (AEDE) e a estimação de um modelo de defasagem espacial (SAR) via Método dos Momentos Generalizado (GMM), buscou-se identificar o comportamento e os determinantes para o repasse federal às microrregiões brasileiras, considerando o componente espacial. O diagnóstico espacial indicou que microrregiões que receberam maiores montantes de repasse eram vizinhas de microrregiões que também obtiveram repasses mais elevados. Resultados econométricos apontaram que microrregiões mais pobres receberam maiores repasses para a saúde. Pelo lado da demanda, a alta complexidade foi priorizada sobre a baixa; o inverso pelo lado da oferta, indicando certo grau de ambiguidade na distribuição dos recursos. Morbidade, número de médicos, hospitais de baixa complexidade e componentes político-partidários também foram determinantes para a atração de recursos federais.
Article
The Brazilian government shares the responsibility of financing public health among federal, state, and municipal levels. Health expenditures are thus uneven across the country and cannot contribute equally to health outcomes across disease categories. This study aims to identify how the health expenditures of municipalities affect the mortality rate in the state of Paraná by causa mortis. We considered years of life lost for each municipality, the chapters of the International Classification of Diseases (10th revision), and the elasticity of this measure in relation to public health expenditure. Considering the possibility of endogeneity, this study follows the instrumental variable approach in a panel of generalized method of moments - instrumental variable (GMM-IV) with fixed effects. Our results show that a 1% increase in health expenditure could decrease the average number of years lost specifically for some causes from 0.176% to 1.56% at the municipal level. These findings could elucidate policy perspective within state finance.
Article
This paper studies the existence of international health outcome disparities. We focus on the use of the potential years of life lost for a database that includes information from 33 OECD countries and covers the period 1990–2017. The methodology proposed by Phillips and Sul (2007) allows us to reject the existence of a single pattern of behaviour between countries for both males and females, suggesting the existence of severe health outcome inequalities. This methodology estimates the existence of four convergence clubs whose composition slightly varies when comparing the male and female cases. Some socioeconomic factors are found to be very important in explaining the forces that may drive the creation of these convergence clubs. In particular, the evolution of the economy and health policies are pivotal to understanding the creation of these estimated convergence clubs. Additionally, our results offer evidence in favor of the importance of environmental policies to explain these health outcome differences.
Article
Social Determinants of Health (SDH) are important in explaining why some countries enjoy better health than others. This empirical study highlights controversies in the literature on the relationship between socioeconomic development and health, and investigates how the relationship might vary in countries with different economic growth. The sample consists of 172 countries, and recent cross-sectional data was collected from the World Bank’s “Data Bank”. Population health was proxied with life expectancy, infant mortality and under-five mortality, and sociooeconomic conditions were expressed with GNI/capita, unemployment rate, poverty, Gini index, educational attainment, public spending on health and physician density. Countries were categorized into three groups as low / lower middle-, upper middle- and high-income according to World Bank thresholds. Ordinary least squares regression was used to model the relationships. The association between socioeconomic conditions and health differed between countries of different economic development. Poverty, educational attainment, income inequality, and physician density were the strongest contributors to health. Higher economic development had a stronger relationship with health in richer countries, and government commitment to health care in poorer ones. Based on evidence from studies such as this one, researchers and policymakers globally could commit to acting together on SDH, and to aligning resources from different sectors to formulate interventions aiming to improve population health.
Article
Full-text available
An Erratum has been published for this article in Health Economics; 14(2): 117 (2005). Canadian per capita drug expenditures increased markedly in recent years and have become center stage in the debate on health care cost containment. To inform public policy, these costs must be compared with the benefits provided by these drugs. This paper measures the statistical relationship between drug spending in Canadian provinces and overall health outcomes. The analysis relies on more homogenous data and includes a more complete set of controls for confounding factors than previous studies. Results show a strong statistical relationship between drug spending and health outcomes, especially for infant mortality and life expectancy at 65. This relationship is almost always stronger for private drug spending than for public drug spending. The analysis further indicates that substantially better health outcomes are observed in provinces where higher drug spending occurs. Simulations show that if all provinces increased per capita drug spending to the levels observed in the two provinces with the highest spending level, an average of 584 fewer infant deaths per year and over 6 months of increased life expectancy at birth would result. Copyright