Xiaomei Lu

Spring 2019

Master thesis 1, 15 ECTS

Master’s in Economics

Determinants of health care

expenditure in Sweden

Xiaomei Lu

Abstract

Sweden faces increasing pressures on health funding. Total expenditure on health

care currently accounts for about 10.92% of GDP, which suggests an increase of about

twofold over the last five decades. This paper examines the short-run and long-run

relationship between income and health care expenditure in Sweden during the period

1980–2017. The study focused on the differences between short- and long-term

elasticities. Consistent with the conventional findings, the income elasticity for health

care is found to be greater than one, suggesting that health care is a luxury good in

Sweden. Additionally, the age structure variable is found to have a significant positive

impact on health care expenditure. Finally, the importance of another non-income

variable, relative price, is also confirmed, an increase in relative price is associated with

lower quantity of health care expenditure.

Table of content

Ⅰ. Introduction ................................................................................................................ 1

Ⅱ. Theoretical bases ....................................................................................................... 3

Ⅲ. Data and analytical framework .................................................................................4

Ⅳ. Econometric methodology and results..................................................................... 5

Ⅳ.Ⅰ Unit root and stationarity tests.......................................................................…. 5

Ⅳ.Ⅱ Co-integration test ..............................................................................................6

Ⅳ.Ⅲ Error Correction Model ......................................................................................7

Ⅴ. Summary and conclusions ........................................................................................9

Acknowledgement........................................................................................................10

References....................................................................................................................11

Appendix......................................................................................................................13

List of tables

Table 1. Unit root test .....................................................................................................6

Table 2.OLS regression on lnqt.......................................................................................6

Table 3. Estimation results, error correction models ......................................................8

Graph The change trend of quantity of total health care expenditure, per capital GDP,

relative price and percentage of population above 65years old during 1980-2017 in

Sweden ...........................................................................................................................5

1

Ⅰ Introduction

The objective of this article is to examine the short- and long-run effects of income

and prices of health care expenditure in Sweden, using 1980-2017 annual time-series

data.

A series of studies1 have focused on using cross-sectional data and panel data to

investigate the relationship between the health care expenditure and the determinants

across countries. These studies differ in terms of the time period covered, determinates

selected, country chosen and econometric techniques employed in the estimation.

Obviously, these studies provide meaningful and useful guide to understand the reason

why health care expenditure varies significantly across countries, but on the other hand,

they cannot be generalized to individual country since the aggregate level empirical

analyses are less able to capture the complexity and peculiarity of the economic

environments of each individual country. Hence, any inferences derived from these

studies can only provide a general understanding of the broad relationship between

these variables.

In this spirit, country-specific case studies may give more exact description on how

income and prices affect health care expenditure in the studied countries. Gbesemete

and Gerdtham (1992); Ang (2010); Seshamani and Gray (2004); Murthy and Okunade

(2016) and Morgan, Gmeinder and Wilkens (2017) examine the evolution and

determinants of health care expenditure in specific country by using systematic time-

series data. To some extent, the main finding of these studies offers guidance and

constructive suggestions for policy formulation in the country-specific studied.

The literature about the relationship between health spending and GDP have been

popularized by the early work of professor Joseph Newhouse (1997). He inferred that

over 90 percent of the sample variation between countries in HE could be explained by

variations in per capita GDP, with an income elasticity in the range 1.15 to 1.31.

Subsequently, other economists have been attracted to the study of international

comparisons of health expenditure and empirical studies2 have supported the hypothesis

that there is strong positive relationship between per capita health spending and per

capita GDP.

1.see e.g. Newhouse (1977), Hitiris (1997), Cuyler (1989), Gerdtham and Löthgren

(2000), Gerdtam and Jönsson (1991), Gerdtham and Johannesson (1999), Hitiris and

Posnett (1992), Lago-Peñas, Cantarero-Prieto, and Blázquez-Fernández (2013),

Mladenović et al (2016), Narayan, Narayan, and Mishra (2010), Braendle and

Colombier (2016), Lorenzoni and Koechlin (2017),etc. 2. see e.g. Gerdtham and Jönsson (1991), Milne and Molana (1991), Hitiris (1997),

Cuyler (1989), Gerdtham and Löthgren (2000), Lorenzoni and Koechlin(2017),

Baltagi and Moscone (2010), Blomqvist and Carter (1996), etc.

2

Other than the GDP per capita, economists are also interested in the non-income

determinants of health care expenditure. A group of studies3have found that the age

structure of the population (usually measured by the percentage of population over age

65) are significantly associated with health care expenditure.

In addition to GDP and the age structure of the population, the relative price of health

care can also be viewed as an determinant that affect the demand of health care.4 In the

literature, Lorenzoni and Koechlin (2017) argued that “health-specific price levels play

an important role in explaining differences in per capita health care volumes across

countries. Hospital price levels for 2014 are for the first time available for all OECD

countries (http://stats.oecd.org/Index.aspx?DataSetCode=PPP2014).” Considering

relative price of health care is related to the quantity of health care funding, such non-

income variable may also be treated indirectly as determinants of health care.

As mentioned above the aim of this article is to examine the short- and long-run

relationship between the health care expenditure and its determinants in Sweden by

using 1980-2017 annual time-series data. In studying the relationship between the

quantity of health care expenditure (qt), the real income(GDP per capita/consumer price

index) (yt) , the relative price (pt) of health care and demographics (OLDt), defined as

the percentage of the population 65 years old and older, this study applied unit root test

and co-integration theory to analyze the econometric relationship between time series

and then adopt the error correction model to unravel the short and long-run relationship

among the variables.

The study’s objective was achieved in three steps. Firstly, non-stationarity of each

observed time-series was tested by using the standard approaches which are to estimate

augmented Dickey–Fuller ADF regression (ADF test) and Phillips–Perron test(PP), and

then ascertain the order of integration of the data series. Secondly, by applying Engle-

Granger two-step method, results were obtained regarding the existence of long-run

cointegration relationships between the variables in the data. Thirdly, the short and

long-run elasticities of real income (yt) and relative price (pt) were estimated by

adopting an Error Correction Model (ECM).

Briefly foreshadowing the main conclusions, the present study finds that the quantity

of health care expenditure(qt), the real income (yt), the relative price (pt) of health care

and demographics (OLDt) are integrated of order one and are cointegrated in the Engle-

Granger cointegration test. The long-run results reveal that while real income (yt) and

demographics (OLDt) have a positive and statistically significant effect on the quantity

of health care (qt), relative price (pt) is also statistically significant but with a negative

effect as expected. More specifically, the current study finds that the short-run elasticity

of quantity of health care with respect to real income is below one, around 0.81-0.88,

while the long-run elasticity is above one, in the range of 1.42-1.64, which suggest that

health care in long-run is a luxury good in Sweden.

3see e.g. Hitiris and Posnett (1992), Lago-Peñas, Cantarero-Prieto, and Blázquez-

Fernández(2013), Ang (2010), Braendle and Colombier (2016).

4see Milne and Molana(1991), Leu (1986), Gerdtham and Jönsson (1991b), Gerdtham

and Jönsson (1991a) .

3

This paper is most closely related to Ang (2010) who studied the determinants of

health care expenditure using Australian data from 1960 to 2003 and found that the

determinants, including real income, demographic structure (both percentage of

population above 65years old and below 15years old), density of health care services

and public finance on health care expenditure, impose an a significant positive influence

on health care expenditure. Specifically, pointing out the long-run elasticities of health

care expenditure with respect to real income are found to be in the range of 1.207–1.252

implying that health care is luxury good in Australia. The present paper contributes by

introducing in the relative price which is assumed to have an impact on the quantity of

health care, and using relatively new Swedish data.

The rest of this article is organized as follows: Section II analyses the theories on

which this paper is based. In section III, the data and analytical framework for

modelling was described. In section Ⅳ the econometric techniques employed in this

study are described, and also the results are presented and analyzed. The last section

summarizes and concludes.

Ⅱ Theoretical bases

The paper is based on a few fundamental economic theories.

The theory of consumer demand is applicable for explaining the relationship between

the health care expenditure and real income, also the relative price of health care. The

income effect is a phenomenon observed through changes in purchasing power,

revealing the change in quantity demanded brought by a change in real income. In

general, the relationship between income and quantity demanded is positive, while

classic economics predicts that quantity demanded decreases with increasing price.

As we know, normal goods are those goods for which the demand rises as income

rises, while inferior goods are goods whose demand decreases as consumer income

rises (or demand increases when consumer income decreases). (Jurion, B.J. 1978)

Obviously, health care appears to be a normal good. On the other hand, a luxury good is

a good for which demand increases by a greater proportion than income. (Varian,

Hal 1992) In contrast, a necessary good is a good for which demand increases by a

smaller proportion than income. For the long-run elasticity is found to be above one, in

the range of 1.41-1.64, which suggests that health care in long-run is a luxury good in

Sweden.

For most goods, elasticities are often lower in the short run than in the long run.

Changes that just are not possible to make in a short amount of time are realistic over

a longer time. In the short-run, the health care expenditure is lagging behind GDP

growth. Adjustments for demanding and supplying need a certain period of time to be

realistic. For instance, the increase in health care centers, medical personnel and

hospital beds etc. takes time to adjust after observing the increasing in the economic

growth.

4

Ⅲ Data and analytical framework

In the previous health economics studies, per capita GDP, the proportion of

population under 15 and/or over 65 have been identified as the key determinants of

health care expenditure. Other than these two determinants, this study introduces the

relative price of health care into the analytical framework, since economic theory

predict that the relative price of a good should affect the quantity for a normal good.

The formulation of the empirical specification of the health care demand equation

was given in Equation 1.

lnqt=ƒ(lnyt, lnpt, OLDt) (1)

where qt is the ratio of total health care expenditure per capital to the producer price

index PPI(2015=100) of medical and dental instruments and supplies; yt is the ratio of

per capital GDP to the consumer price index CPI（1980=100）; pt is the producer price

index (2015=100) of medical and dental instruments and supplies divided by the

consumer price index CPI; and OLDt is the percentage of population aged over 65.

All variables (other than OLDt) are expressed in natural logarithms for the usual

statistical reasons. The reason for using producer price index (2015=100) of medical

and dental instruments and supplies instead of the consumer price index of medical care

to derive the relative price is that the household out-of-pocket payment of the total

health care expenditure account for a small proportion, around 15%. However, up to

56% of the total health care expenditure in Sweden is funded by the county councils

and enterprises owned by the county councils and about 25% is funded by

municipalities. In such particular circumstance, the consumer price index does not fully

reveal the relative price trend of the health care, but the producer price index is an

appropriate alternative choice. Annual data covering the period 1980–2017 in Sweden

were used in the estimation. All data were gathered from OECD Health Data (2018)

and SCB, (except that the PPI during the period 1980-1990 is estimated on CPI 1980-

1990 through OLS regression, because of the lack of data resource for it, as explained

in the appendix.)

A graph of the changing trend of quantity of the total health care expenditure (lnqt),

per capital GDP (lnyt), the relative price (lnpt) and the percentage of population above

65years old (OLDt) is presented as follow. In general, the growth rate of quantity of the

total health care expenditure (lnqt) and per capital GDP (lnyt) is similar as predicted,

while the relative price (lnpt) and the percentage of population above 65years old (OLDt)

appears comparatively constant.

5

Ⅳ Econometric Methodology and results

Ⅳ. Ⅰunit root and stationarity tests

Before adopting any time series model, stationary tests for all variables are necessary

in order to detect and avoid spurious regressions. A unit root test is used to study if a

time series variable is non-stationary and possesses a unit root. In general, there are

several tests for it, including augmented Dickey–Fuller test(ADF), Phillips–Perron test,

KPSS test, ADF-GLS test. In this article, augmented Dickey–Fuller test(ADF) and

Phillips–Perron test(PP), which are the most common unit root test, were used to test

the stationarity of the time series. With this aim, ADF test and PP test for lnqt, lnyt, lnpt

and OLDt have been carried out. A summary of results is reported in Table 1. The null

hypothesis (H0) that the tested series has a unit root is accepted in the level of all series

which indicates that lnqt, lnyt, lnpt and OLDt are all non-stationary. This results are the

same as Gerdtham and Löthgren (2000) reported for health expenditure and GDP.

However according to both of ADF test and PP test, all series of lnqt, lnyt, lnpt and OLDt

are stationary process after first order difference. Hence all series may be treated as I(1).

-4

-2

0

2

4

6

1980 1985 1990 1995 2000 2005 2010 2015

lnqt lnyt

lnpt OLD1

The change trend of quantity of total health care expenditure, per capita GDP,

relative price and percentage of population above 65years old

during 1980-2017 in Sweden

6

Table1 unit root test

lnqt lnyt lnpt OLDt

test types(c t m) t-statistic t-statistic t-statistic t-statistic

ADF test

(0,0,0)

(c,0,0)

(c,t,0)

(0,0,1)

(c,0,1)

(c,t,1)

1.33

0.483

-1.99

-4.7***

-4.84***

-5.41***

1.32

0.39

-1.69

-3.33***

-3.59***

-4.19***

-1.39

-2.23

-2.88

-6.42***

-6.65***

-6.58***

0.72

-2.03

-3.68**

-3.14***

-3.23***

-4.24***

PP test

(0,0,0)

(c,0,0)

(c,t,0)

(0,0,1)

(c,0,1)

(c,t,1)

1.09

0.19

-1.99

-4.74***

-4.82***

-5.41***

1.30

0.51

-1.63

-3.24***

-3.59**

-4.22***

-1.23

-2.05

-2.80

-6.69***

-7.09***

-6.98***

1.89

-0.21

-0.99

-4.20***

-5.69***

-8.67***

Notes: For ADF, AIC was used to select the lag length and the maximum number of

lags was set to be nine. For PP, AR spectral-GLS detrend was used as the spectral

estimation method. *, ** and *** denote significance at the 10%, 5% and 1% levels,

respectively. For the test type (c,t,m) , c denotes with intercept, t with intercept and

trend, and m indicates lagged differences.

Ⅳ.Ⅱ Co-integration test

Since all series are stationary process after first order difference, there might exist

cointegration relationships between the variables. By applying the Engle-Granger two-

step method, results can be obtained regarding the existence of long-run cointegration

relationships between the variables in the data.

In the first step, OLS regression was conducted with lnqt as the explained variable

and lnyt, lnpt and OLDt as the explanatory variables. The regression result is

presented as follow (Table2):

Table 2 OLS regression on lnqt

Variable Coefficient Std. Error t-Statistic Prob.

C -6.14 0.65 -9.396 0.0000lnyt 1.42 0.082 17.22 0.0000lnpt -1.23 0.36 -3.41 0.0017

OLDt 7.72 2.18 3.55 0.0012

R-squared 0.958 Mean dependent var 3.468Adjusted R-squared 0.955 S.D. dependent var 0.304S.E. of regression 0.064 Akaike info criterion -2.54Sum squared resid 0.142 Schwarz criterion -2.367Log likelihood 52.25 Hannan-Quinn criter. -2.478F-statistic 261.64 Durbin-Watson stat 0.73Prob(F-statistic) 0.000000

7

In the second step, the residual series is made out from the OLS regression above

and then conducted the ADF test for the residual series generated. The t-statistic value

of the residual series in the ADF test is -2.86, less than the critical value of the 1%

significance level, -2.63. Therefore, it can be considered that the residual error series

is stationary, indicating the existence of long-run cointegration relationships between

the variables in the data.

Ⅳ.Ⅲ Error Correction Model

Since the residual of the OLS regression is considered to be stationary in the

cointegration analysis, which indicates the existence of long-run cointegration

relationships between the variables. The error correction model (ECM) can be adopted

to estimate the long-run equilibrium relationship between quantity of health care

expenditure and its determinants. By applying the ECM, the short-run dynamic

adjustment has also been revealed, which are critical in any time-series investigation.

The effects of the determinants of health care expenditure are often different in the short

and long run.

Below the error correction model (ECM) was set out, by just formulating one

equation Δlnqt=α0+α1Δlnyt+α2Δlnpt+α3ΔOLDt+λlnqt-1+β1lnyt-1+β2lnpt-1+β3OLDt-1+еt

(referred as model A) and by using Engle-Granger two-step approach (referred as

model B), respectively.

In model A, the short-run elasticities of all the determinants can be obtained

directly from the estimated result, including the real income elasticity, the price

elasticity and demographic semi-elasticity. The corresponding long-run elasticities

can be derived indirectly by adopting βK/-λ, for k=1,2,3. When applying this model,

the error correction term is not necessary to be generated during the regression

process.

On the other hand, while adopting model B, it is necessary to obtain the error

correction term first, by regressing the long-run steady-state model in the first step

and making residuals out of it. The long-run elasticities are obtained directly during

first step. While in the second step, by bringing in the error correction term, error

correction model (ECM) is set out and short-run elasticities can be obtained.

Firstly, the following long-run steady-state solution is obtained in the co-

integration test. lnqt=-6.14+1.42 lnyt-1.23 lnpt+7.72 OLDt (refer to Table 2)

Next, the Error-Correction Term (ECT) can be obtained by taking lnqt-1-a0–b1lnyt-1

- b2lnpt-1 +b3OLDt-1 to formulate the error correction model (ECM) as shown in

Equation Δlnqt=δ1Δlnyt+δ2Δlnpt+δ3ΔOLDt-μ(ecm t-1)+еt .Where ecmt-1 = lnqt-1-a0–

b1lnyt-1 - b2lnpt-1 +b3OLDt-1. The ECT captures the health care expenditure evolution

process by which agents adjust for prediction errors made in the last period. The

stationarity of the ECT provides evidence that the long-run relationship exists.

Slightly different results are acquired from the two models. The results are

presented in Table 3. But in general, all long-run effects are significant at the

conventional levels and have the expected signs and reasonable magnitudes.

8

Table 3 Estimation results, error correction models

Model A

Model B

(Engle-Granger two-step approach)

Variable parameter coefficient Std.error Prob. parameter coefficient Std.error Prob.

Panel 1 the long-run equilibrium level relationship

c

lnyt

lnpt

OLDt

β1 /-λ

β2 /-λ

β3 /-λ

1.64

-3.07

12.79

0.31

1.68

6.97

0.000

0.068

0.067

a0

b1

b2

b3

-6.14

1.42

-1.23

7.72

0.65

0.08

0.36

2.18

0.000

0.000

0.001

0.001

Panel 2 the short-run relationship

c

Δlnyt

Δlnpt

ΔOLDt

lnqt-1

lnyt-1

lnpt-1

OLDt-1

ecmt-1

α0

α1

α2

α3

λ

β1

β2

β3

-2.81

0.81

-1.34

-2.91

-0.27

0.45

-0.84

3.49

1.02

0.24

0.34

8.4

0.13

0.19

0.38

1.98

0.009

0.001

0.0004

0.731

0.045

0.029

0.038

0.089

δ1

δ2

δ3

μ

0.88

-1.01

7.58

-0.303

0.21

0.25

4.46

0.126

0.000

0.000

0.099

0.021

Model A

Model B(1-step)

Model B(2-step)

R-squared

0.62

0.958

0.58

Adjusted R-squared

0.52

0.955

0.54

F-statistic

6.66

261.64

Prob(F-statistic)

0.000

0.000

S.E of regression

0.046

0.06

0.045

(1) The regression results for model A, reported in panel 2 of Table 3, provide the

short-run dynamic changes pattern. Other than the demographic variable (OLDt) which

indicates no short-run inference can be made about this variable, all other coefficients

are statistically significant at the 5% level. In the first differenced form, real income(lnyt)

and relative price (lnpt), appear to be statistically significant at the 1% level, and

therefore short-run inference can be made. The short-run income elasticity is 0.81,

indicating when per capital GDP grows 1%, the quantity of health care will increase by

0.81% in the current year. The short-run price elasticity has the expected negative sign

and implies that an increase in relative price of health care by 1% will decrease the

quantity consumed by 1.34%.

The coefficient on lnyt-1 is about 0.45, indicating the quantity of health care in the

previous year increasing by 1% will lead to the quantity in current year rising by

0.45% .The possible explanation is that the fixed assets and personnel investment

proposed in the previous year in health care have a stimulating effect on the health care

expenditure in the following period of time.

The long-run elasticity of quantity of health care with respect to per capita GDP is

around 1.64 with standard error 0.31, suggesting that the real income long-run elasticity

is above 1. The result implies that health care in long-run is a luxury good in Sweden.

9

Such a finding is consistent with a number of previous studies, including Newhouse

(1977), Leu (1986), Gerdtham and Jönsson (1991), Milne and Molana (1991), Hitiris

(1997) and Ang (2010). The long-run price elasticity is around -3.07 with standard error

1.68. This point estimated is below that reported by Gerdtham and Jönsson (1991),

Marc Pomp and Sunčica Vujić(2008), in which the relative price elasticities are between

minus one and zero. However due to the large standard error, the 95% confidence

interval range from -6.43 to -0.29, and therefore do not contradict that the true effect is

between minus one and zero. The possible explanation is that the use of predicted prices

for some years resulting in the large standard error. And the use of different approaches

to derive the relative price of the health care could be another reason.

(2) Engle-Granger two-step approach (model B)

The regression results for model B are presented in table 3.

The coefficients on ecmt-1 is -0.303, which measure the speed at which adjustment

returns to the long-run equilibrium value, is statistically significant at the 1% level

with the expected negative sign. This indicates that the quantity of health care adjusts

at the speed of 30.3% every year to restore equilibrium when there is a shock to the

steady-state relationship.

While the estimated short-run income elasticity δ1 is around 0.88, the long-run

elasticity is 1.42 with standard error 0.08. This result also suggests that health care in

long-run is a luxury good in Sweden. And while the estimated short-run price elasticity

δ2 is around -1.01, the long-run elasticity is -1.23 with standard error 0.36. The

magnitudes of the coefficients for relative price are slightly different from that obtained

in model A, and closer to that found in previous studies.

The coefficient on OLDt is found to be around 7.58 with the expected (positive)

sign and statistically significant at the 1% level. The result suggests that the quantity

of health care expenditure rises significantly with the increase in elderly population

(over 65) relative to total population, which is expected. To this extent, the results

support the view that demographic factors are crucial in explaining the variations in

health care expenditure, as reported by the previous studies of Gerdtham and Jönsson

(1991a),Hitiris( 1997) ,Gerdtham et al. (1992) and Murthy and Ukpolo (1994)

Ⅴ Summary and conclusions

Sweden faces increasing pressures on health funding. Total expenditure on health

care currently accounts for about 10.92% of GDP in Sweden, suggesting an increase of

about twofold over the last five decades. It is crucial to examine the underlying causes

of this increase and provide some fundamental guidelines for health policy designing.

This paper investigated the short and long-run economic relationship between

quantity of health care and its determinants, including per capital GDP, relative price

and age structure (the percentage of population above 65years old). Using a time-series

data of Sweden followed about 39 years, this paper has studied the non-stationarity and

cointegration properties of all the series. The analysis indicates that health care

expenditure and the three determinants studied are non-stationary, but there exists long-

10

run cointegration between the variables. Our results show that there exists strong

positive relationship between quantity of health care and per capita GDP as reported in

previous studies, and also indicate that health care in long-run is a luxury good in

Sweden, with a long-run income elasticity greater than one, consistent with the previous

studies Newhouse(1977). As for non-income determinants, our analysis indicates the

percentage of population above 65yeas old plays a quite important role in explaining

health expenditure variations as previous studies implied. And finally, from the results

obtained, relative price is also an important determinant of the demand for health care

since its coefficient is significant at 1% level in both models.

Several limitations of this study need to be acknowledged. One limitation concerns

the fact that there may be some omitted variables which affect the demand for health.

This could lead to omitted variables bias. Another limitation concerns issues of data.

Because of lacking for relative price from 1980 to 1990, it had to be imputed. On the

other hand, future studies on health care expenditure growth may further expand the

study by empirically test different variables and related health indexes with GDP. And

it may be worthwhile to try using longer time series to study if the elasticities have

changed over time.

Acknowledgement

In the acknowledgement of this thesis, I would like to thank different people who

directly or indirectly contributed to my thesis via their time, efforts, knowledge,

expertise. Particularly I would like to thank my supervisor, David Granlund, for his

efforts, support and guidance. David has provided his feedback and supervision

throughout the research work in a professional manner. His criticism and suggestions

were always constructive as far as the improvements in the thesis were concerned. I

also would like to thank those who supported me throughout the process.

11

References

Newhouse P. Joseph (1977), Medical-Care Expenditure: A Cross-National Survey, The

Journal of Human Resources, Vol. 12, No. 1 (Winter, 1977), pp. 115-125

Gerdtham ULF -G, Jes Sørgaard, Fredrik Andersson, and Bengt Jönsson (1991) , An

econometric analysis of health care expenditure: A cross-section study of the

OECD countries. Journal of Health Economics I1 (1992) 63-84 North-Holland

Gerdtham ULF -G and Bengt Jonsson (1991), Price and quantity in international

comparisons of health care expenditure. Applied Economics, 1991, 23, 1519-1528

Gerdtham ULF -G and Jonssön Bengt (1991), Conversion factor instability in

international comparisons of health care expenditure. Journal of Health Economics 10

(1991) 227-234. North-Holland

Gerdtham ULF -G and Johannesson Magnus (1997), New estimates of the demand for

health: results based on a categorical health measure and Swedish micro data.

Working Paper Series in Economics and Finance No. 205 November 1997

Gerdtham. Ulf-G and Löthgren Mickael (2000), On stationarity and cointegration of

international health expenditure and GDP, Journal of Health Economics 19 2000 461–

475

Lago-Peñas, Cantarero-Prieto, Blázquez-Fernández (2013) On the relationship

between GDP and health care expenditure: A new look. Economic Modelling 32 (2013)

124–129

Morgan David, Gmeinder Michael and Wilkens Jens (2017) An OECD analysis of

health spending in Norway. OECD Health Working Papers No. 91

Culyer, A.J (1989), Cost Containment in Europe. Health Care Financ Rev. 1989 Dec;

1989(Suppl): 21–32.

Gbesemete Kwamep and Gerdtham Ulf-G (1992), Determinants of Health Care

Expenditure in Africa: A Cross-Sectional Study. World Development Vol. 20. No. 2. pp.

303-30X. 1992 .

Murthy N.R. Vasudeva, Okunade A. Albert (2016), Determinants of U.S. health

expenditure: Evidence from autoregressive distributed lag (ARDL) approach to

cointegration. Economic Modelling 59 (2016) 67–73

Baltagi H. Badi and Moscone Francesco (2010) Health care expenditure and income

in the OECD reconsidered: Evidence from panel data. Economic Modelling 27 (2010)

804–811

Ellis P. Randall, Martins Bruno and Zhu Wenjia (2017), Health care demand

elasticities by type of service. Journal of Health Economics 55 (2017) 232–243

Hitiris. Theo (1997), Health care expenditure and integration in the countries of the

European Union, Applied Economics, 29:1, 1-6, DOI: 10.1080/000368497327335

Hitiris. Theo and Posnett John (1992), The determinants and effects of health

expenditure in developed countries, Journal of Health Economics 11 (1992) 173-18 1.

North-Holland

12

Getzen E. Thomas (2000), Health care is an individual necessity and a national

luxury: applying multilevel decision models to the analysis of health care

expenditures.Journal of Health Economics 19 2000 259–270

Mladenović Igor, Miloš, Milovančević, Svetlana Sokolov Mladenović , Vladislav

Marjanović, Petković (2016) Analyzing and management of health care expenditure

and gross domestic product (GDP) growth rate by adaptive neuro-fuzzy technique,

Computers in Human Behavior 64 (2016) 524e530

Lorenzoni Luca, Koechlin Francette (2017) , International Comparisons of Health

Prices and Volumes: New Findings, OECD Directorate for Employment, Labour and

Social Affairs.

Blomqvist, Å.G and Carter. R.A.L (1997), Is health care really a luxury? Journal of

Health Economics 16 (1997) 207-229

Abdul Azeez Oluwanisola Abdul Wahab and Zurina Kefeli (2016), Projecting a Long

Term Expenditure Growth in Healthcare Service: A Literature Review, Procedia

Economics and Finance 37 ( 2016 ) 152 – 157

Gerdtham.Ulf-G and Löthgren Mickael (2000), On stationarity and cointegration of

international health expenditure and GDP, Journal of Health Economics 19 2000 461–

475

Schieber J. George and Poullier Jean-Pierre (1989), Overview of international

comparisons of health care expenditures, Health Care Financ Rev. 1989 Dec;

1989(Suppl): 1–7.

Leu, R. E. (1986) The public–private mix and international health care costs, in Public

and Private Health Services (Eds) A. J. Culyer and B. Jonsson, Basil Blackwell,

Oxford, pp. 41–63.

Murthy N. R. Vasudeva & Ukpolo Victor(1994), Aggregate health care expenditure in

the United States: evidence from cointegration tests. Applied Economics Volume 26,

1994-Issue 8

Milne R and Molana H (1991), On the effect of income and relative price on demand

for health care: EC evidence. Applied Economics (1991) 23(7) 1221-1226

Seshamani Meena and Gray Alastair (2004), Ageing and health‐care expenditure: the

red herring argument revisited. Health Economics Volume13, Issue4 April 2004 Pages 303-

314

Ang B James (2010), The determinants of health care expenditure in Australia, Applied

Economics Letters, 17: 7, 639 — 644, First published on: 01 June 2009 (iFirst)

Braendle Thomas and Colombier Carsten (2016), What drives public health care

expenditure growth? Evidence from Swiss cantons, 1970–2012, Health Policy 120

(2016) 1051–1060

Wang Kuan-Min (2011), Health care expenditure and economic growth: Quantile

panel-type analysis, Economic Modelling 28 (2011) 1536–1549

Narayan Seema, Narayan Kumar Paresh and Mishra Sagarika (2010), Investigating

the relationship between health and economic growth: Empirical evidence from a

panel of 5 Asian countries, Journal of Asian Economics 21 (2010) 404–411

13

Liang Li-Lin and Mirelman J.Andrew (2014), Why do some countries spend more for

health? An assessment of sociopolitical determinants and international aid for

government health expenditures, Social Science & Medicine 114 (2014) 161e168

Duarte Fabian (2012), Price elasticity of expenditure across health care services,

Journal of Health Economics 31 (2012) 824–841

Zhou Zhongliang, Su Yanfang, Gao Jianmin, Xu Ling, Zhang Yaoguang (2011), New

estimates of elasticity of demand for healthcare in rural China, Health Policy 103

(2011) 255–265

Appendix

Data of the consumer price index and producer price index was collected from

SCB. While the CPI is presented from 1980 to 2017, the PPI only presented from

1990 to 2017. Considering that the PPI generally appears to have similar trend as CPI.

PPI in the period of 1980-1990 is predicted by using data of CPI.

Firstly, set out regression equation lnPPI=α+βlnCPI, following results were obtained.

lnPPI Coef. St.Err. t-value p-value [95% Conf Interval]

lnCPI 1.054 0.065 16.23 0.000 0.921 1.188

Constant -1.520 0.365 -4.17 0.000 -2.270 -0.770

Mean dependent var 4.403 SD dependent var 0.126

R-squared 0.910 Number of obs 28.000

F-test 263.561 Prob > F 0.000

Akaike crit. (AIC) -100.859 Bayesian crit. (BIC) -98.194

*** p<0.01, ** p<0.05, * p<0.1

Next, PPI was predicted by using the reported regression model. PPI during the period

of 1980-1990 is obtain then.

The data of total health care expenditure and GPD is collected from OECD health

care database.